Nagent FAQ: AI teammates, approvals and budgets

Nagent is an agentic execution platform, not a chatbot or copilot. Copilots assist: they generate suggestions, drafts, or answers that humans still need to act on. Nagent goes a step further by deploying autonomous AI agents that actually execute workflows across your business. These agents don't just respond to prompts. They:

• Plan tasks
• Take actions across systems
• Coordinate with other agents
• Continuously improve outcomes

This means your teams move from manual execution → autonomous execution, with humans shifting to oversight and strategy.

Nagent can automate end-to-end workflows across functions, not just isolated tasks. Examples include:

• Marketing: campaign creation, content production, ad generation
• Sales: lead qualification, call analysis, pipeline tracking
• Operations: reporting, research, coordination workflows
• Hiring: candidate sourcing, outreach, market mapping

Unlike traditional automation, Nagent agents handle multi-step, decision-driven workflows that combine data, logic, and actions across tools.

No. Nagent is built for growth teams, not engineering teams.

• Brief an AI teammate such as MIRA, SERA or CREA in plain language, and approve its work before it ships.
• Engineers can extend the platform through APIs and integrations when they want to.

People and AI agents work in one shared context, so nobody has to build an agent to get the work done.

Most enterprise teams deploy their first agent within 48 hours, and scale to multiple workflows within weeks. Nagent accelerates deployment through:

• Pre-built agents and templates
• Agentic AI Lab (for guided implementation)
• Ready integrations with existing tools

The goal is not experimentation. It's a fast transition to production execution.

Nagent offers AI teammates for each part of growth work, including:

• MIRA for marketing, with CREA for content, RUPA for images, Alpha for video and HOOK for paid social creative
• SERA for sales and NORA for outbound research
• DRIS for search and answer engine visibility, and NIA for Google Ads
• Echo for reputation and FOLIO for long-form publications

Each teammate works at a named autonomy level, and a person approves its work until it earns the right to act alone.

Nagent drives measurable impact by:

• Reducing manual work across teams
• Accelerating execution speed
• Increasing output quality and consistency
• Enabling continuous optimization

Typical outcomes include: faster campaign deployment, higher content velocity, improved conversion rates, and reduced operational overhead.

You can get started in three ways:

• Pick the AI teams you need, such as Marketing with MIRA or Sales with SERA
• Brief a teammate with the number you want to move
• Book a demo for a guided rollout

For enterprises, we also offer an Agentic AI Lab to identify high-impact use cases and deploy production-ready agents quickly.

Yes. Nagent is designed specifically for enterprise-scale deployment. It supports:

• Multi-team usage
• Governance and access control
• Scalable infrastructure
• Integration with complex systems

This makes it suitable for organizations looking to move from experimentation to organization-wide AI adoption.

Nagent stands out in three key ways:

• Execution first: agents do not just assist, they carry work through to pipeline and customers
• One memory, one record and one ladder of trust: every agent reads the same context, writes every action to an audit record and earns autonomy as its record grows
• Enterprise readiness: security, control, and scalability built in

While many platforms focus on building agents, Nagent focuses on making them work in production at scale.

Nagent is multiplayer AI for growth teams: people and AI agents in one shared context, delivering the number. A company's people and its AI teammates work across marketing, sales and content in team rooms, from the same brief, the same record and the same next move.

• AI teammates are led by chiefs of staff: MIRA for Marketing, SERA for Sales and CREA for Content.
• Agents answer from the company's own Knowledge Hub, with sources, and say so when the record does not hold the answer.
• Agents earn autonomy level by level on their track record, and a person approves anything that goes live.

A general assistant works for one person at a time: the context lives in that person's chat history, it knows nothing about your company, and nobody else can see what it did. Nagent puts your people and a team of AI agents in one shared context, grounded in your own knowledge, with every action proposed, approved, run and audited.

Agents answer from your Knowledge Hub with sources. Each agent sits on a level from L1 to L5 and earns more room through approved work. Publishing and destructive changes always wait for a person.

Those are libraries for engineers: toolkits to wire agents together. With them, your team builds the memory layer, the governance gate, the audit log, the observability stack, the deployment surface, the integration plumbing, and the operator UI yourselves. Twelve months and an ML team later, you have what Nagent ships out of the box.

Those are conversational AI products built for one function, customer support. Nagent is built for growth teams: AI teammates for marketing, sales and content working beside your people in one shared context, grounded in your own knowledge, with approvals and budgets a person can see.

Nagent is built with enterprise-grade security and data sovereignty at its core. Key capabilities include:

• Deployment options: cloud, VPC, or on-prem
• Role-based access control (RBAC)
• Full audit logs of every agent action
• Data residency controls
• No training on your proprietary data

Nagent is ISO/IEC 27001:2022 certified by InterCert, and a SOC 2 Type II programme is underway.

Yes. Nagent is designed to work seamlessly with your existing stack. Agents can integrate with:

• CRM systems (e.g., Salesforce, HubSpot)
• Marketing tools
• Internal databases
• APIs and custom systems

This allows agents to operate within your current workflows, rather than replacing your infrastructure.

800+ tools via Composio (CRMs, email, chat, data warehouses, marketing automation) plus native deep integrations for the platforms where we need more control: Anthropic, OpenAI, Cal.id (native meeting layer), ContactOut (outbound discovery + enrichment), Exa, Storyblok, Razorpay, Cloudinary.

Yes, with Brand Workspaces. Each workspace is fully isolated: its own brand book, its own agent governance, its own model preferences, its own audit trail. Operators can switch between workspaces with a click; permissions are scoped per-workspace. Built for agencies and multi-brand enterprises.

Sonnet 4.6 is the default for most production work. It hits roughly 98% of Opus 4.7's capability at one-fifth the price. Pick Opus 4.7 when the task is autonomous multi-file engineering or where complexity meaningfully rewards the price step. Pick Haiku 4.5 for high-volume routing, classification, or low-stakes summarisation.

It lets Claude operate a browser or desktop app directly (clicking, typing, reading the screen) without a custom integration per app. Excellent for legacy SaaS automation. Accuracy still varies on complex flows, so wire it through Nagent's human-in-the-loop approval queue for high-stakes work.

Repeated context (system prompts, tool schemas, large reference documents) is cached for 5 minutes. Cache reads cost roughly 10% of cache writes. For typical agent loops where the same context recurs every turn, total spend drops by 60 to 80% with caching enabled.

Nagent pools across multiple Anthropic API keys and tiers, so typical workloads see no rate limiting. During burst spikes the router auto-fails over to OpenAI / Mistral as configured fallbacks, so your agents keep responding.

Opus 4.7 wins on multi-file engineering and natural prose quality. GPT-5.5 wins on broad general-purpose reasoning, vision tasks, and is the practical default if you also need o-series chain-of-thought reasoning in the same workflow.

When the task needs explicit chain-of-thought, such as STEM problems, scientific reasoning, multi-step logic, financial modelling. For everyday agent loops (drafting, classification, tool-use) GPT-5.4 is faster and cheaper.

Yes: gpt-realtime-1.5 is wired into our voice-agent skill alongside Cartesia and ElevenLabs. Pick OpenAI when you want a single provider for both reasoning and speech in the same agent.

Yes for high-stakes outputs where the cost of a wrong answer dwarfs token cost, such as legal review, medical documentation and regulated workflows. Use it sparingly through Nagent's tier-aware router so cheaper models handle the routine subtasks.

When your agent needs current facts or attributed answers. Gemini queries Google's live index in-line and returns inline citations. Adds slight latency but eliminates the need for a separate web-search tool in many agent stacks.

No. Nano Banana 2 and Pro are Google's reasoning-first image models built on the Gemini family. Best when image accuracy matters: text rendering, real product likeness, character consistency across a series.

Use Gemini 3.1 Flash for routine 1M-context work ($0.14 / $0.28 per 1M tokens). Reserve 3.1 Pro for complex reasoning that genuinely needs the full window. Flash-Lite handles high-volume simple tasks at ~$0.025 per 1M.

Yes. Gemini supports EU data residency through Vertex AI. Configure the region in your Nagent workspace settings before deploying so all inference and storage stay on EU infrastructure.

Yes, especially for competitive intel, news monitoring, and any workload that touches current events. It cuts the need to maintain a separate web-search or news-API tool for most use cases.

You can pass a video URL or upload a clip directly; Grok reasons over it without a separate transcription step. Useful for analysing call recordings, demo videos, or training content in one step.

It doesn't. That's internal to Grok. From an integration standpoint Grok is a normal LLM endpoint; you simply get measurably better creative and analytical chat outputs.

xAI is pricing aggressively to win share. Output quality matches frontier providers on most benchmarks; the trade-off is the platform is younger and the ecosystem (third-party tools, fine-tuning options) is smaller than OpenAI's.

Sonar is right for high-volume cited chat. Sonar Pro is right for production answer engines where retrieval depth and reasoning quality both matter. Pro costs roughly 3x more on output tokens but the answer quality jump is significant.

Largely yes, for initial drafting and source aggregation. One call produces what a junior analyst would deliver in 5 to 15 minutes. For final synthesis and judgement calls, pair the output with a human reviewer.

Citations come back as structured metadata on the response. Nagent's action score ledger persists them so you can audit any factual claim back to its source URL, useful for compliance reviews and regulated reporting.

Yes. Many agents use Perplexity as a tool call from a Claude or GPT brain. The LLM decides when to call Sonar for grounded retrieval, then synthesises the answer using its own reasoning. Best of both worlds.

Two reasons: EU AI Act compliance (data must remain on EU infrastructure) and air-gapped self-hosting (open weights). For everything else, the closed-model frontier is currently stronger on raw capability.

Your model weights and inference run on infrastructure you control rather than the model provider's servers. This is required for some regulated industries, most defence work, and a growing list of EU-public-sector tenders.

On 80+ language coverage and 256K context for codebase-wide refactors, yes. For interactive single-file work, GPT-5.3 Codex still has an edge. Use Nagent's router to pick per-task automatically.

Yes. Our tier-aware router can default to closed-model frontier for hard tasks and fall back to Mistral when data residency requires it. The agent definition stays the same; the router handles provider selection.

Yes for input tokens at frontier-quality output. Verify on your actual workload using our side-by-side cost dashboard. The savings vary by prompt mix, but the order of magnitude is real.

Opus still wins on long-horizon multi-file changes and complex architectural reasoning. DeepSeek wins on per-file changes, code review, routine refactors, and bug fixes, all at a fraction of the per-call cost.

For most commercial agent work, no. It carries the same risk profile as other providers using web-scale training data. For high-IP-sensitivity use cases, talk to us about self-hosting the open weights inside your VPC.

Yes. They all run on standard inference stacks (vLLM, TGI, SGLang). Nagent's sovereign-deployment supports the full open-weight roster on a single Kubernetes cluster with shared GPU pools.

Yes for commercial deployment. Apache 2.0 lets you ship Qwen inside customer products with no further licensing. Mistral Large 3 currently requires a research license, which adds friction for embedded/OEM scenarios.

Strong CJK coverage and good EMEA-language quality. The right choice if your agent fleet operates across regions and you would otherwise need per-language model swaps to maintain quality.

Yes for compute planning: 122B needs ~244GB VRAM at FP16, 32B needs ~64GB. On output quality, 32B is the sweet spot for most enterprise agent workloads. The 122B step gives diminishing returns outside specialised tasks.

Yes for self-hosted deployments. Bring a LoRA adapter or do a full fine-tune; Nagent's training infrastructure supports both Qwen base and instruct variants on shared GPU clusters.

Ideogram trains explicitly on text-as-design rather than text-as-noise. The result is letterforms that survive professional design review, with accurate kerning, consistent stroke weight, brand-correct typeface execution.

Yes, roughly $0.04 to $0.08 per image depending on tier. Subscription plans ($8 to $60/month) are for the consumer interface; for automated pipelines you pay per generation through the API.

Yes via system prompts and reference images. Nagent's Campaign Hub agent encodes a brand style guide once and applies it consistently across all generations: colours, fonts, layout grid, photography style.

Ideogram for typography-led work: posters, logos, ad creative, anything where text-on-image accuracy matters. Nano Banana for context-grounded accuracy: real product likeness, factually correct scene composition.

Only when image quality directly drives revenue: luxury hero assets, magazine spreads, OOH campaigns, premium brand work. For most volume use cases, Ideogram or Nano Banana 2 is the right cost/quality answer.

More painterly control, finer photographic detail, stronger style fidelity to specific art-direction references. The difference is visible to design reviewers; less visible to general consumers in scroll-by social contexts.

Yes for hero shots, but Seedream 5.0 from ByteDance is more cost-effective for batch product variations. Nagent's router can split: Flux for the hero, Seedream for the variant grid.

Yes via the standard API. Black Forest Labs is German-based; data residency is generally EU-friendly. Confirm specific data flows with your legal team if you handle regulated data.

OpenAI hasn't published a definitive reason. Industry analysis points to the "physics-first" approach being computationally expensive, with high audio failure rates (reportedly ~75%). Veo 3.1's joint audio/video architecture proved superior on both cost and quality.

Migrate now if you have ongoing pipelines. Nagent's migration wizard re-points workflows to Veo 3.1 or Kling 3.0 in one config change. Waiting risks last-minute pressure when the API actually shuts off.

Yes for cinematic quality and joint audio. For social and short-form video, Kling 3.0 is actually a better fit: multi-shot narrative control plus roughly 10× cheaper per second. Pick based on use case, not as a 1:1 swap.

Yes as of May 2026. Other models top out at 1080p; upscaling to 4K loses physics consistency. Veo's unified architecture renders 4K natively without that quality loss.

Audio and video are produced from the same generation pass, so sounds match the visual physics, such as water, fabric, footsteps or clinking glass. Other workflows generate video and audio separately and then stitch them, which produces noticeable mismatches.

Cheaper than even low-budget production but more expensive than Kling 3.0 for non-premium use cases. Use Veo when broadcast-quality matters; use Kling when social-first and cost-per-clip matters more.

Yes for individual hero clips. For multi-clip narrative ads, generate 8-second clips and stitch them. Nagent's video-pipeline agent handles the assembly, audio sync, and transitions automatically.

Mid-tier on price and quality. Best for product animation and lifestyle content where you need 3D consistency but don't need Veo's 4K + audio premium. Cheaper than Veo, slightly more expensive than Kling.

You provide the first and last frame as images; Luma interpolates the in-between motion. Excellent for narrative continuity in product showcases or branded content where the start/end states are fixed.

Better than the legacy Sora 2; not as strong as Kling 3.0's multi-shot mode. If character continuity is the primary requirement, Kling is the right pick.

Kuaishou prices for volume. They expect to make margin on social-content creators generating dozens of clips per day rather than premium hero assets. The unit economics target a different buyer.

A single prompt generates multiple connected clips with consistent characters and scene continuity. Replaces the need to chain three separate generations together and stitch them. The model handles the continuity.

Yes for social and short-form work. For broadcast-quality cinematic, Veo 3.1 is still preferred. Use action scores on your actual workload to validate quality before committing to a full pipeline.

Yes when paired with Cartesia or ElevenLabs voice. Kling handles motion + lip-sync; the voice provider handles audio. Nagent's video-pipeline agent stitches the two outputs in one workflow.

You can input up to 9 images, 3 videos, and 3 audio clips as references for a single video generation. It's the highest level of director control on the market: the closest thing to a creative brief that the model actually respects.

Possibly in Western markets long-term. Nagent's failover routing to Veo / Kling is automatic if Seedance becomes unavailable in your region, so your agents continue running without code changes.

Seedream for product batches: the same product across many lifestyles or environments while keeping the product itself perfectly consistent. Ideogram for typography-heavy design work. They're complementary, not competing.

Inference runs on whichever endpoint you configure. For Western deployments use the US/EU endpoints; for Asia-Pacific the regional endpoints. Confirm specific data flows with your compliance team before deploying regulated workloads.

ElevenLabs for media production where audio quality is the primary requirement, such as audiobooks, dubbing and narration. Cartesia for real-time conversation where end-to-end latency dominates user perception of quality.

You provide 60 minutes of clean source audio; the clone is typically ready within 24 hours. The result is indistinguishable from the original in blind A/B tests at conversation length.

You need consent from the voice owner. ElevenLabs and Nagent both require attestation before activating PVC. Use cloned voices for owned brand assets only. Never use them to imitate a real person without explicit consent.

Eleven Flash v2.5 hits ~75ms model latency, fast enough for most interactive use. For sub-100ms requirements (call centres, live conversation), Nagent's router falls back to Cartesia automatically.

Cartesia uses State Space Model (SSM) architecture instead of transformers. The math is fundamentally faster for streaming audio generation. That's where the 40ms model latency comes from.

Close, not equal. ElevenLabs still wins on emotional range and naturalism for long-form work. For real-time conversation, the latency benefit dominates the small remaining quality gap. Users perceive responsiveness as quality.

Yes via voice cloning, with a simpler workflow than ElevenLabs PVC: just a few minutes of source audio. Quality is good for conversational use; for long-form narration, ElevenLabs PVC is still the better choice.

Yes. It is a drop-in replacement for ElevenLabs in our voice-agent skill. Pair it with any LLM brain (Claude, GPT-5, Grok) for the reasoning side. The agent definition doesn't change; just swap the speech provider.

Pricing model. Smallest AI's flat $0.05/min PAYG fits contact-centre unit economics cleanly. Cartesia's subscription + credits model is harder to forecast for high-concurrency workloads where call volume varies dramatically.

Strong English coverage and decent multilingual. For broad multilingual scale at predictable cost, Minimax is a better fit. For English-only telephony at the lowest cost, Pulse is the most cost-efficient choice we offer.

Yes. It pairs with our outbound-sales agent. The combination of low latency and low cost makes outbound voice economically viable at scale, especially in markets where per-call margins are thin.

Both target telephony. Lightning is optimised for higher-concurrency burst patterns, such as peak-time call centres where thousands of calls land in the same minute. For most workloads Pulse is the default.

When you need broad multilingual coverage at predictable PAYG pricing without per-language subscriptions. For single-language English work, ElevenLabs (quality) or Cartesia (latency) win.

$60 per 1M characters is a flat meter, which is easier to forecast than per-second voice subscriptions when traffic patterns are unpredictable. Particularly useful for content TTS where the input is text, not stream length.

Fine for most conversational use, noticeably slower than Cartesia's 40ms in side-by-side high-stakes real-time. Acceptable for content TTS, voice notifications, and async voice work where the user is not actively waiting.

Yes. It is particularly good for multilingual podcast localisation at predictable cost. For premium-quality original podcast production in a single language, ElevenLabs remains the standard.

They work from your own material. The Knowledge Hub holds your website, documents and pasted notes; the customer profile describes who you sell to; and offerings list what you sell once a person confirms them.

• Point Nagent at your website and it reads your key pages into the Knowledge Hub.
• Add more as pasted text, a URL or an uploaded document; each is indexed so agents can search it.
• In a team room, replies are built from the team's record: charter, shared documents, decisions, tasks and agreed memory, each with sources.
• When nothing relevant is found, the agent says it cannot cite an answer and asks for the source.

An honest miss is cheaper than a confident wrong answer, so an uncited figure is discarded rather than posted.

Through a record of approved work, one level at a time. Every agent action is logged, and every approval is evidence. Each night Nagent nominates an agent for the next level once it has spent 30 days at its level with at least 25 decided approvals, fewer than 5 in every 100 of them dismissed, and a workspace admin approves or declines.

A promotion is offered, never automatic, and a reversal resets the record. Destructive changes ask a person every time, at every level.

Every Nagent agent sits on one of five levels: L1 Observe, L2 Assist, L3 Operate, L4 Represent and L5 Lead. Each action an agent can take declares the level it needs to run without a person:

• Pure reads always run.
• When the agent's level meets the action's bar, the action fires.
• When it does not, an approval request is written and nothing runs until a person decides.
• Destructive changes ask a person every time, at every level.

Beyond per-agent levels, you can flag per-action overrides (e.g. "any outbound to a C-level requires approval even at L4"). One kill-switch click pauses an agent, an agent class, or the whole fleet.

For high-stakes calls, such as whether a $50k email batch should go out, whether a lead should be upgraded to a hot opportunity or whether to recommend a discount, Nagent runs a moderated debate between specialist agents (Decision Engine, Cost-Cap, Brand-Voice, Risk Assessor) before any action fires.

Round-robin cadence with a hybrid stop condition (consensus or N rounds). $20/day per-debate spend cap. Full debate transcript logged. Every other agentic platform on the market today ships a single-agent execution model.

A company signs up and picks its departments: Sales, Marketing or Content. Each department installs a team into the workspace, a lead agent and the agents who work under it, seated in a team room. The set-up chat then asks for what the team needs, such as your website, your brand and the tools to connect.

Instructions in a room become tasks on the team's board only after a person confirms them, and every agent carries a daily and a monthly spending cap.

Topology is the real-time cockpit of every agent in your fleet: what each one is doing right now, what it just decided, where the next handoff is queued, which agents are debating, which are waiting on approval.

Operators can pause, demote, isolate, or override any agent from the cockpit without losing the run state. Most agentic platforms ship a chat box; Nagent ships the cockpit underneath it.

A team is installed on your first sign-in: pick Sales, Marketing or Content and the lead agent and its teammates are seated in a team room. The set-up chat then asks for what the team needs, such as your website, brand and tools, so you can brief the team in the first hour.

Every agent action is logged and attributed to the business outcome it produced: replies, meetings booked, deals closed, revenue won, hours saved. The /admin/insights/conversion funnel shows per-source, per-owner, per-ICP-band attribution. Beta cohort metrics: 10× faster campaign cycle (FMCG, 12-week pilot), 60% operational overhead removed (aggregate across 8 customers).

On the Enterprise plan, yes: private VPC deployment, bring your own model and your own key, with custom guardrails and approval. Every plan has tenant isolation, RBAC and an append-only audit trail. A BFSI governance profile adds a stricter approval floor for regulated teams.

Three append-only logs back the platform:

• AuditLog: every write op, 365-day retention, immutable hooks.
• AgentActionLog: every agent run with resolution and error context, 90-day retention.
• ExecutionContext: context-evaluation snapshots, 30-day retention.

Plus the Decisions Ledger: every AI-suggested action logged against the entity it affects with the outcome it produced. Operators can replay the chain to understand why a particular decision was made.

Not yet. Nagent is ISO/IEC 27001:2022 certified, with a SOC 2 Type II programme underway. Immutable AuditLog, RBAC across 7 roles, JIT access via approval engine, MFA for sensitive roles, device-trust gates, kill switch, full per-agent action history.

Seven kinds, configurable per agent without touching the model layer:

• Cost cap: daily / per-run / per-debate spend ceilings.
• Content restriction: banned phrases, topic guardrails.
• Approval required: flag specific actions to queue for human review.
• Audience restriction: limit which contacts / segments the agent can act on.
• Tool restriction: narrow the action surface the agent can compose from.
• Time window: only fire during allowed hours / days.
• Custom: operator-defined predicates.

Each plan is a monthly or annual subscription with seats and a monthly allowance of credits, the unit of agent work. The plans, their prices and their credits are on the pricing page.

Tiered platform fee + usage-based metering for model spend, with hard per-agent and per-day caps so spend never surprises you. The Enterprise plan adds private VPC deployment, bring your own model and key, and custom guardrails and approval.

See /pricing for tier comparison, or book a 30-minute working session for a tailored proposal.

Nagent supports multi-agent orchestration, where multiple agents collaborate to complete complex workflows. For example:

• One agent gathers data
• Another analyzes it
• A third executes actions

This coordinated system allows Nagent to handle end-to-end processes, not just individual steps.

MIRA leads Nagent's AI Marketing team. You brief her with the outcome you want, and she:

• Plans the work against that outcome
• Hands each piece to the AI teammate who owns it, such as CREA for content, RUPA for images or Alpha for video
• Brings the work back to your team for approval before it ships

Your people and the AI teammates share one context, so the plan, the drafts and the results stay in one place.

MIRA, SERA and CREA lead the Marketing, Sales and Content teams. NORA researches outbound accounts, DRIS measures AI search visibility, NIA runs Google Ads, HOOK writes ad creative strategy (Beta), Echo watches reputation (Beta), RUPA renders images, Alpha plans and renders video, and FOLIO turns source documents into long-form publications.

Every teammate works in the shared workspace, grounded in your Knowledge Hub and under the same approvals and budgets. See /platform.

Stacks are multi-agent compositions: the output of one agent feeds the next, with conditional branches, parallel fan-out and handoffs. Each stack carries its own memory, its own track record and its own governance bounds.

Skills are reusable markdown prompt modules attached to one or more agents. Each defines when to use it (trigger description) and how to apply it (the prompt body), with optional allowedTools[] to restrict which actions it can compose.

Operators can create skills from prose, from a PDF upload, or by hand on the /admin/skills page.

79 production templates seeded from the GTM design system, each tagged with persona, funnel stage, tone, and category. EmailBuilder generates and ranks variants for a given lead context; the operator picks one; each choice is recorded so the variants that win are ranked higher over time.

Browse at /admin/email-templates. The "Seed 79 GTM templates" button on the same page re-syncs from the bundle.

NORA is Nagent's AI outbound research lead. She finds accounts that fit your customer profile, researches each into a sourced dossier and drafts the first e-mail for a person to approve. She is cost-capped, and on a new workspace every outbound send waits for a person.

MIRA is the AI chief of staff for the Marketing team. Brief her with the outcome and she plans it, hands each piece to the teammate who owns it and reports what the agents did and what is waiting. See /mira.

On Nagent, the AI agent for lead enrichment is NORA. She finds accounts that match your customer profile, qualifies them and researches each into a dossier with sources. Research and contact reveals run against a daily outreach budget. She reads the workspace's own lead records, and research and drafting run on their own while every send waits for a person.

NORA does not stop at filling fields. Each account becomes a sourced dossier, the dossier feeds the outreach she drafts, and she reads every reply until a person takes the meeting. She works in the Sales team room beside your sellers, and every action she takes is written to the action log with its trigger, cost and resolution.

NORA works from your saved customer profile, the workspace's own lead records and the Knowledge Hub, which holds your brand, buyers, rules and results. Outside systems are reached through connectors granted one key at a time, and every tool action is ticked on individually before any agent may use it.

NORA comes with the Sales team. Choosing Sales at sign-up installs the team into your workspace, seated in one team room, and a named engineer works with you through onboarding. From there you save the customer profile she works against, and her research starts from it.

Volume is something you set rather than something left open-ended. Saved campaigns carry a cap on profiles per run, research and contact reveals run against a daily outreach budget with a cost-cap exception queue, and every agent has a daily and a monthly spending cap. That keeps research spend predictable as the list grows.

Partly. NORA reads the workspace's own lead records. An outside CRM can be connected as a tool, under a connection only your workspace can use, but her pipeline does not sync records with it. Every tool action is ticked on individually before any agent may use it, so you decide exactly what she can touch.

NORA's dossiers carry their sources, so a seller can check where each finding came from, and her drafts only make claims about the reader that trace to something she found. Drafts wait on her Messages surface to be read, edited, approved or rejected, so a person sees the work before anything reaches a prospect.

On Nagent, an AI agent for CRM hygiene is SERA's Sales team, which keeps the deal record in order. SERA reports the pipeline from the CRM digest, never invents a deal, a contact, a value or a close date, and says what is missing where the record is silent. Each fix is handed to an owner, and changes wait for a person.

Not at first. Every agent starts with every action waiting for a person, and every tool action on an outside CRM is ticked on individually before any agent may use it. An agent earns more room one kind of action at a time, and destructive changes ask a person every time.

SERA works from the CRM digest in the team record. An outside CRM can be connected as a tool, under a connection only your workspace can use, through a connector granted one key at a time. NORA's pipeline does not sync records with it, and every tool action is ticked on individually.

A cleanup tool changes records by rule. On Nagent, people and agents work in one team room on the same record: SERA names what is missing, an instruction becomes a task only when a person confirms it, approvals wait on cards, and every action is logged with its trigger, target, cost and resolution.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room with SERA leading it. If the record holds no pipeline yet, SERA says so and names who can import it. A named engineer works with you through onboarding and a named operator through delivery.

SERA does not fill gaps: where the CRM is silent she says what is missing, and she never invents a deal, a contact, a value or a close date. For new prospects, NORA researches each account into a dossier with sources, and research and contact reveals run against a daily outreach budget.

Decisions are kept. A decision recorded in a room is permanent and the team's agents can cite it back, and an approval decided on a card in the thread or the Inbox is decided once. Each agent's memory sits on top of the Knowledge Hub and keeps what worked and what failed.

On Nagent, an AI agent for sales enablement is the Sales team in its room. SERA answers where deals stand from the CRM digest and names the next step for a stuck deal. LIPI drafts e-mail to leads already in the CRM, VANI drafts NORA's outreach, and a person approves before anything is sent.

Each Nagent agent has a job, a manager and a record. SERA works in a team room beside your sellers, hands work only to teammates on the roster, and answers from the CRM record rather than from memory. Every action is fired, queued for a person or refused, and written to the action log.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room with SERA leading it. Adding your offerings and documents to the Knowledge Hub gives every agent the same source to read. A named engineer works with you through onboarding and a named operator through delivery.

Reps work with the team in one room. Deals reach it once they are in the CRM, and CRM e-mail goes out from the rep's own Gmail or Outlook mailbox, so the sender is the person who owns the lead. Google Calendar supplies free and busy times and the events a booked meeting creates.

Client data stays in the client's tenant, every workspace is isolated and every role is scoped. A member who may not read leads gets the stage roll-up from SERA and never a named deal. Every agent action is written to an audit record the client can read, and Enterprise can deploy in a private cloud or on premise.

They write from what the workspace holds. LIPI and VANI write in the workspace's voice from its offerings and knowledge only, so each offering the workspace records is material they can draw on, and nothing outside it. Every e-mail stays a draft until a person approves it, because a sent e-mail cannot be unsent.

It goes to a person. Before any agent acts, Nagent fires the action, queues it for a person or refuses it, and where the record is silent SERA says what is missing instead of guessing. A queued action waits as a card in the thread or the Inbox, and every resolution is logged.

On Nagent, the AI agent for outbound sales is NORA. She finds accounts that match your customer profile, researches each into a dossier with sources, drafts email outreach, reads the replies and hands a qualified prospect to a person. She reports to SERA, the Sales chief of staff, and a person takes every meeting.

NORA works from research rather than a fixed script. Each draft comes from the dossier she built for that prospect, and each reply is classified so the cadence stops or changes course. She works in a team room beside your sellers, sends wait for a person, and every action is logged with its trigger, cost and resolution.

NORA comes with the Sales team. Choosing Sales at sign-up installs the team into your workspace, seated in one team room, and a named engineer works with you through onboarding. Her first campaign starts from the customer profile you save, and on a new workspace every send waits for a person.

Drafts are written per prospect from the research dossier NORA built and from your workspace's own knowledge and offerings, so every claim about the reader traces to something she found. VANI writes the first touch, follow-ups and breakup email in your workspace's voice, and a person reads each draft before it is sent.

NORA reads every reply and classifies it as interested, objection, scheduling or unsubscribe, with anything else set aside as noise. A qualified reply becomes a meeting and a handoff card on her Meetings board. A named person acknowledges it, and a handoff nobody acknowledges within 36 hours escalates to the top of the board.

NORA reads the workspace's own lead records. An outside CRM can be connected as a tool, under a connection only your workspace can use, but her pipeline does not sync records with it. Every tool action is ticked on individually before any agent may use it, and SERA answers pipeline questions from the CRM digest.

Each saved campaign carries a cap on profiles per run, and research and contact reveals run against a daily outreach budget and a cost-cap exception queue. Every agent also has a daily and a monthly spending cap. One campaign runs from list to handoff, sourcing, drafting, approval and the handoff, in one window.

On Nagent, the AI agent for pipeline management is SERA. Ask in the Sales room and she reports open deals per stage, the value in each stage and the age of the oldest from the CRM digest. For a stuck deal she names the next step and who owns it. She never invents a deal, a contact, a value or a close date.

Automation rules fire on a trigger. SERA works in a team room with your sellers, answers pipeline questions from the record in the thread, and hands work to the teammate who owns it. The control plane fires, queues for a person or refuses every action, and writes each one to the action log.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room, and SERA leads it. Every Nagent agent is a configuration of ten declared parts rather than a program you write. A named engineer works with you through onboarding and a named operator through delivery.

SERA answers from the CRM digest in the team record. An outside CRM can be connected as a tool, under a connection only your workspace can use, with every tool action ticked on individually, but NORA's pipeline does not sync records with it. CRM e-mail goes out from the rep's own Gmail or Outlook mailbox.

A person approves at first. Every agent starts with every action waiting for a person. When a class of action reaches 20 approvals with no reversals, Nagent raises one promotion request and a person grants or declines it. Destructive changes ask a person every time, and sending is never started from a room.

SERA hands forecasting to the teammate on the Sales roster who owns it. RIVA, the revenue operations lead who keeps the pipeline and the forecast honest, is coming soon. Today SERA answers from the CRM digest, stage by stage, and says what is missing wherever the record is silent.

For a question about one deal, SERA reads up to 25 open deals with stage, owner, value and last activity, and named deals reach only members who may read leads. Client data stays in the client's tenant, every role is scoped, and every agent action is written to an audit record the client can read.

On Nagent, the AI agent for competitor monitoring is DRIS. He tracks how ChatGPT, Claude, Gemini, Perplexity and Grok answer the questions your buyers type, scoring share of voice and recommendation position among six dimensions, alongside competitor tracking, ranked keywords, Search Console and traffic. Competitor ads are studied by HOOK's SCOUT, who saves them to a swipe library with the reason each was kept.

An alert returns links. DRIS asks a fixed set of buyer questions, records the exact prompt behind every citation, and scores each target the same way before and after. Every gap becomes one entry in a ranked queue with the reason, the fix and the confidence attached, and every figure carries the run count, the window and the engine.

Nagent does not quote a set time. DRIS comes with the Marketing team, which installs as one unit: a roster, a lead and a room. A person sets his objective and baseline and connects Search Console and Google Analytics 4, which he reads daily. A named engineer goes with you through onboarding.

DRIS covers search and answer engines: five answer engines on one fixed prompt set, ranked keywords, Search Console, traffic and competitor tracking. Competitor ads belong to HOOK, whose SCOUT finds the ads worth studying and whose MINE works out what is working and why. Echo listens to reviews and mentions about your own company, not about competitors.

DRIS does not send a stream of alerts. Every gap goes into one ranked Queue with its reason, the fix and the confidence, and reports say what moved against the baseline. A person works the Queue in order, and approving an item clears a step to run; it does not publish anything.

Yes, as a baseline. DRIS measures every target the same way before and after, and the prompt set is fixed per account so a trend means something. The exact prompt behind each result is recorded alongside its citation, and every executed fix carries the actor and timestamp in the audit trail, inside your tenant.

You do not build DRIS. He arrives on the Marketing team already knowing the job, as one of the teams Nagent ships rather than a blank canvas. Search Console and Google Analytics 4 connect through connectors granted one key at a time, a named engineer goes with you through onboarding, and Enterprise plans can deploy in your private VPC.

On Nagent, the AI agent for social media management is not one bot. MIRA plans the work, CREA writes posts, RUPA makes the images, Alpha makes social video and Echo's SOCIAL seat watches the threads a person adds. Every action that needs approval waits for a person, and a person posts. MOXA, the dedicated social media lead, is coming soon.

No. You brief the team in plain words in a team room: address an agent by name and it answers in the thread, and a message addressed to nobody goes to the team lead. An instruction becomes a task on the team board only when a person confirms it. A named engineer goes with you through onboarding.

CREA holds every factual claim to the knowledge, offerings or documents the workspace has shared, and the voice comes from the brand book, which stays a draft until a person confirms it. RUPA's Brand Lock is on by default, so the brand guidelines go with every render, and the bench shows the exact text sent.

No. No teammate posts to a social network today. RUPA files her images in the Asset Library, nothing is published, scheduled or sent by CREA, and Alpha never publishes. A person posts the approved work. MOXA, coming soon, will lead organic social, with each post approved by a named person before it goes out.

Approvals sit with people throughout. Each action that needs approval waits as a card in the thread, and changes that cannot be undone are never started from a room. On Echo's side, a review about safety, health, discrimination, harassment, a legal threat or a child's wellbeing is graded sensitive, and no seat drafts its reply: the designated reviewer writes it.

Nagent does not schedule posts today, so it is not a scheduler. It is a team: MIRA plans the week, routes each piece to the teammate who owns it and keeps the thread together, while the people who own the number approve the work in the same room. Every agent action lands in the action log with its trigger and resolution.

Human on the loop. Every agent starts with each action waiting for a person, and earns more room one kind of action at a time on its record, with a person signing each promotion. RUPA, for example, can be trusted to render on her own. What reaches a customer stays with people at every level, so posting stays with a person.

On Nagent, the AI agent for fashion product photoshoot brief work is RUPA, and she produces the images rather than a shot-list document. From your brief she renders a sheet of plates, Packshot on sweep or Lifestyle scene, four plates each, followed by a Brand check card. If you need a written brief for a crew, CREA's Content team writes briefs.

Brand Lock is on by default, so the brand book's guidelines go with every render, and the bench shows the exact text sent. With the mark switched on, your logo travels as an image to be placed, never redrawn. A Brand check reads the brief for banned words and marks what text cannot show as Needs your eye.

No. A person writes the brief in RUPA's studio, picks a preset and attaches the product shot. In the Content room, typing @ and picking RUPA gets one brand-locked image from a message, square, wide or tall, standard or HD. A named engineer goes with you through onboarding, and a named operator through delivery.

Each render starts from its own brief, so footwear, outerwear or accessories each get the words, preset and product shot that fit them. The two photoshoot presets are Packshot on sweep and Lifestyle scene, and the studio holds nineteen presets in seven groups. A reference with a stated intent can carry a category's look from one render to the next.

Quality follows what RUPA is given: a fuller brand book, a reference with a stated intent, the product shot attached. Restyle, Relight and Cut out refine a saved plate. Her room renders are logged one by one, and Nagent reads that record to propose more trust, or a step down when people keep overriding her; a person decides.

RUPA renders one sheet per brief, four plates for either photoshoot preset, so a range is a series of briefs. Before anything is sent, her daily and monthly budget, a per-action cap for one batch and the workspace's own limits are checked. Over a limit she refuses and names it, and the plan card shows the estimate first.

Product shots come in as attachments from the Knowledge Hub or the Asset Library, and documents connected through Google Drive are read into the Knowledge Hub. Finished plates are filed in the Asset Library, the Assets tab of the Knowledge Hub, where the team picks them up. Alpha can start a product video from a RUPA keyframe.

On Nagent, the AI agent for edtech student outreach is NORA. She finds accounts that match your customer profile, researches each into a sourced dossier, drafts email outreach, reads the replies and hands a qualified prospect to a person. She is built for business outreach, so she fits an edtech team selling to institutions or employers.

NORA comes with the Sales team. Choosing Sales at sign-up installs the team into your workspace, seated in one team room, and a named engineer works with you through onboarding. Outreach starts from the customer profile you save, and on a new workspace every send waits for a person.

Each draft is written from the research dossier NORA built for that prospect and from your workspace's own knowledge and offerings, so every claim about the reader traces to something she found. Volume is bounded on purpose: saved campaigns carry a cap on profiles per run, and research runs against a daily outreach budget.

Email. NORA does not send messages on social networks; templates for social networks can be written, and a person sends them. Her emails go out through Broadcast, which releases only drafts a person has already approved, inside the sending hours you set.

Prospects who never reply get the cadence VANI writes for NORA: a first touch, follow-ups and a closing breakup email, each waiting for a person's approval on a new workspace. When someone does reply, NORA classifies it as interested, objection, scheduling or unsubscribe, and the cadence stops or changes course.

NORA reads the workspace's own lead records. An outside CRM can be connected as a tool, under a connection only your workspace can use, but her pipeline does not sync records with it. Every tool action is ticked on individually before any agent may use it.

Yes. On a new workspace every send waits for a person, whatever level NORA's seat shows. Before any agent acts, Nagent fires the action, queues it for a person or refuses it, and every action is written to the action log. A person takes every meeting, and a handoff nobody acknowledges within 36 hours escalates.

SERA reports the pipeline from the CRM digest in the team record: stage, count, value and age, each as the record gives it. Salesforce can be connected as a tool under a connection only your workspace can use, and any action on it waits for a person until the agent has earned that kind of action.

Not without a person. Every tool action is ticked on individually before any agent may use it, and every agent starts with every action waiting for a person. Destructive changes ask a person every time, and every action is logged with its trigger, target, cost and resolution.

As a tool, through a connector granted one key at a time, under a connection only your workspace can use. Which actions an agent may take depends on the tool actions you tick on. NORA's pipeline does not sync records with an outside CRM, and SERA answers from the CRM digest in the team record.

Nagent's Sales team works in a team room beside your sellers. SERA answers pipeline questions in the thread from the record, names the next step and the owner for a stuck deal, and hands work to the teammate who owns it. The control plane fires, queues or refuses every action and writes it to the record.

SERA never invents a deal, a contact, a value or a close date, and where the CRM is silent she says what is missing. Changes wait for a person until an agent earns that kind of action, approvals sit on cards in the thread or the Inbox, and every action is logged with its resolution.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room with SERA leading it. If the record holds no pipeline yet, SERA says so and names who can import it. A named engineer works with you through onboarding and a named operator through delivery.

Each part has an owner. SERA hands work only to a teammate on the Sales roster and says in the thread who holds it now. NORA runs outbound research, LIPI writes CRM e-mail in Beta, and DEXA for deal execution and RIVA for revenue operations are coming soon.

It works beside HubSpot rather than replacing it. HubSpot can be connected as a tool under a connection only your workspace can use, but NORA's pipeline does not sync records with it. NORA researches prospects against your saved ideal customer profile, reads every reply and hands a qualified prospect to a person.

As a tool, through a connector granted one key at a time, under a connection only your workspace can use. Each tool action is ticked on individually before any agent may use it. A named engineer works with you through onboarding and a named operator through delivery.

NORA's pipeline does not sync records with an outside CRM, and no agent may use a HubSpot tool action until it has been ticked on individually. Every agent starts with every action waiting for a person, and destructive changes ask a person every time, at every level an agent reaches.

Update the saved ideal customer profile, and NORA finds and researches new prospects against it. The Knowledge Hub is the single source of truth for your brand, buyers, rules and results, and every agent reads it before it works, so the second agent knows what the first one learned.

Only through a tool action you have ticked on, under a connection only your workspace can use. Before any agent acts, Nagent fires the action, queues it for a person or refuses it, and every agent starts with every action waiting for a person. An agent earns more room one kind of action at a time.

NORA's cadence runs a first touch, follow ups and a closing breakup e-mail, which VANI writes. Replies are classified as interested, objection, scheduling or unsubscribe, so the cadence stops or changes course. On a new workspace every send waits for a person, whatever level her seat shows.

Enterprise teams get the same Sales team with enterprise controls: client data stays in the client's tenant, every workspace is isolated, every role is scoped and every agent action is written to an audit record the client can read. Nagent deploys in its own cloud, your private cloud or on premise.

On Nagent, an AI agent for inbound qualification is SERA's Sales team working from the CRM record. Once a lead is in the CRM, SERA reports it from the record, hands the follow-up to the teammate who owns it and says what is missing where the record is silent. A person judges the lead, approves every send and takes the meeting.

Scoring rules produce a number. SERA answers in the team room, from the CRM digest and the named deal records, and says what is missing where the record is silent instead of guessing. She hands each piece of work to the teammate who owns it and says who holds it now, while people decide which leads qualify.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room with SERA leading it. Every Nagent agent is a configuration of ten declared parts, not a program. A named engineer works with you through onboarding and a named operator through delivery.

SERA works from the CRM, so a lead counts once it is a record there, whichever channel it came from. Outside tools can be connected under a connection only your workspace can use, with every tool action ticked on individually before any agent may use it. People then judge every lead from the same record.

A person decides which leads qualify, so no lead is dropped by a rule nobody can see. A lead that is not ready can get a nurture sequence step drafted by LIPI, which is in Beta. Nurture steps wait for a rep to review unless the workspace changes that setting, and every draft comes from the workspace's offerings and knowledge only.

Yes. Every agent action is logged with its trigger, target, cost and resolution, and the room thread says who holds each piece of work. The team record keeps the board and the decision log, and a decision recorded in a room is permanent, so a disputed handoff can be traced back through the record.

That is how the Sales team is built. SERA, her teammates and your sellers work in one thread. A person approves every send, takes the meeting, judges the deal and closes it. A booked meeting becomes a handoff card owned by a named person, and one nobody acknowledges within 36 hours escalates.

A template tool fills variables into a fixed structure. On Nagent, the AI agent for cold email writing works from research: NORA builds a sourced dossier on each prospect, and VANI drafts the emails from that dossier and your workspace's own knowledge and offerings, so every claim about the reader traces to something NORA found.

NORA and VANI come with the Sales team. Choosing Sales at sign-up installs the team into your workspace, seated in one team room, and a named engineer works with you through onboarding. Writing starts once NORA has a saved customer profile and researched prospects to write to.

Yes. VANI writes NORA's first touch, follow-ups and a closing breakup email for prospects who have never replied. NORA reads every reply and classifies it, so the cadence stops or changes course when a prospect shows interest, objects, asks to schedule or unsubscribes.

Each email starts from the research dossier NORA built for that prospect, with sources, and from your workspace's own knowledge and offerings. VANI writes in your workspace's voice, and every claim about the reader traces to something NORA found. A person reads each draft before it is sent.

A person approves them. On a new workspace every send waits for a person, whatever level NORA's seat shows, while drafting and research run on their own. Drafts wait on her Messages surface to be read, edited, approved or rejected, and Broadcast releases only already-approved drafts inside the sending hours you set.

The prospect's research dossier, the workspace's own lead records and its knowledge and offerings, with the Knowledge Hub as the single source of truth for your brand and buyers. An outside CRM can be connected as a tool under a connection only your workspace can use, though NORA's pipeline does not sync records with it.

A writing tool returns a draft and stops. NORA researches the prospect first, VANI writes from that research, and NORA reads the replies until a person takes the meeting. The work happens in the Sales team room beside your sellers, and every action is written to the action log with its trigger, target, cost and resolution.

On Nagent, an AI agent for demo prep gathers the record for the person taking the call. NORA's research dossier with sources, the reply that led to the booking and SERA's read of the deal from the CRM sit in one team room. The meeting becomes a handoff card owned by a named person, who takes the call.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room. Connect Google Calendar for free and busy times and booked events, and Gmail or Outlook so e-mail goes out from the rep's own mailbox. A named engineer works with you through onboarding.

Put them in the Knowledge Hub or pin them to the Sales room, where they become part of the team record that agents read before they work. LIPI and VANI write e-mail in the workspace's voice from its offerings and knowledge only, so the material you add shapes what they write.

Client data stays in the client's tenant and every workspace is isolated. In the Sales room, a member who may not read leads gets the stage roll-up from SERA and never a named deal. Every agent action is logged with its trigger, target, cost and resolution, and Enterprise can deploy in a private cloud.

Yes. The team room is one conversation that a department's people and its AI teammates share, and everything said in it is attributed and kept. Everyone preparing reads the same thread and the same pinned documents, and the handoff card still names the one person who owns the meeting.

NORA's dossier carries its sources, so every claim about a prospect traces to something she found and can be checked. SERA never invents a deal, a contact, a value or a close date, and says what is missing where the record is silent. Feedback in the room changes how an agent works without retraining.

A writing tool starts from a prompt. Nagent's Sales team starts from the record: the saved customer profile, NORA's sourced dossier, the CRM digest and the team record, with the Knowledge Hub every agent reads first. The meeting has a named owner and a 36-hour escalation, which a writing tool does not give you.

On Nagent, the AI agent for account research is NORA. She finds accounts that match your customer profile, qualifies them and researches each into a dossier with sources that a seller can read before a call. The dossier then grounds the outreach she drafts, and on a new workspace every send waits for a person.

An enrichment tool fills fields. NORA produces a sourced dossier per account and carries it through: drafts are written from it, replies are read and classified, and a qualified prospect is handed to a person on a handoff card. She works in the Sales team room beside your sellers, on one shared record.

NORA comes with the Sales team. Choosing Sales at sign-up installs the team into your workspace, seated in one team room, and a named engineer works with you through onboarding. Research begins from the customer profile you save, and research and drafting then run on their own.

NORA reads the workspace's own lead records. An outside CRM can be connected as a tool, under a connection only your workspace can use, but her pipeline does not sync records with it. Every tool action is ticked on individually before any agent may use it, so you decide what she can reach.

Every dossier carries its sources, so a seller can check where a finding came from before acting on it. NORA's drafts make claims about the reader only where they trace to something she found, and each draft waits for a person to read, edit, approve or reject it before it is sent.

Scale is something you set, not something left open. Saved campaigns carry a cap on profiles per run, and research and contact reveals run against a daily outreach budget and a cost-cap exception queue. Each account still gets its own dossier with sources, and every agent has a daily and a monthly spending cap.

The dossier grounds the first touch, follow-ups and breakup email VANI writes for NORA, and on a new workspace every send waits for a person. NORA reads each reply until a person takes the meeting. Once a deal is in the CRM, SERA answers where it stands from the CRM digest and names the next step if it sticks.

An AI agent for content marketing at Nagent is CREA's Content team. Give it a topic and it researches, drafts, edits, reviews, illustrates and links a blog post, then waits for a person. CREA also writes briefs, articles, e-mails, posts and scripts, checked against the brand, and passes images to RUPA and long-form publications to FOLIO.

A general-purpose writer fills gaps with confident generalities. CREA's team grounds every draft in your brand profile, products, Knowledge Hub and brand book, works in a team room your people share, and hands work to the teammate who owns it. Every action is governed, logged and held for approval until the agent has earned more trust.

No. The Content team arrives already built, with CREA leading it and RUPA, FOLIO and Alpha as team leads under her. You brief it in plain words in the Content team room, or type a topic and an optional angle in Blog Studio. Every Nagent agent is a configuration of ten declared parts, not a program you write.

The voice comes from your brand book, and a voice extracted from your material stays a draft until a person confirms it. The same style rules go to the writer and the reviewer. The Knowledge Hub is the single source of truth for the brand, and product names come from Products and services, so they are never invented.

Yes, and the last step always belongs to a person. An action that needs approval waits as a card in the thread or the Inbox, and deciding in either place decides it once. Publishing a post stays with a person until someone explicitly grants it, and under the stricter governance profile, content always asks.

CREA writes briefs, articles, e-mails, posts and scripts, and the team produces illustrated, linked blog posts in Blog Studio. RUPA makes images such as ad creatives, social posts and product photoshoots. FOLIO turns your documents into long-form publications as PDF and DOCX, and Alpha makes video for the Content team. A presentation maker and a landing page maker are coming soon.

Through the record and the room. Feedback from the room changes how an agent works without retraining, and each agent's memory keeps what worked and what failed. DRIS re-scores visibility against the baseline it started from and hands new content gaps to the Content board as tasks, so the next brief starts from what is measured.

An AI agent for blog writing on Nagent is CREA's Content team. Given a topic, it fetches citations, drafts in your brand voice, polishes, reviews, generates a cover and inline visuals, adds internal links and runs an SEO audit of the title, meta description and readability. Then it waits: a person approves and publishes.

A general chat assistant writes from what it already knows. CREA's team writes from your brand profile, your products and the Knowledge Hub passages that match the topic, appends a Sources section, links only to posts and entities your site already has, and runs a review and an editorial check. The final publish stays with a person.

No. The Content team comes with the plan that includes it, already structured with CREA as its lead. In Blog Studio you type a topic and optionally an angle, and the run page shows the work as it happens. Every agent is a configuration of declared parts rather than a program anyone writes.

The writer and the reviewer receive the same house style rules, and the voice comes from your brand book. A voice extracted from your material stays a draft until a person confirms it. Product names come from Products and services, so they are never invented, and every factual claim must trace to a retrieved passage.

Part of it. The SEO audit checks the title, meta description and readability, and the team applies entity links and a related posts section using only what your site already has. At Approve you can edit the title, meta description and slug yourself. DRIS covers wider organic visibility and hands content gaps to the Content board.

Published posts live in the workspace, in Posts, where every post, AI and human, is filtered by status. Your site reads published posts from the workspace through the site connection, which is in Beta and uses its own site key. Google Drive documents are read into the Knowledge Hub the team writes from.

That is set by a daily blog budget rather than a fixed count. Each Blog Studio run is checked against the budget before it starts, and when the budget is spent a new run stops until the next day. The Content Studio shows what the day has spent, and every agent also carries its own spending cap.

On Nagent, an AI agent for expansion playbooks works from the record. SERA answers where each account's deals stand from the CRM digest and names the next step and owner for a stuck one. LIPI drafts e-mail to leads already in the CRM. A person approves each send, takes the meeting and decides the deal.

A workflow fires on a rule. SERA works in a team room beside your people, answers from the CRM record and says what is missing where it is silent, and hands work only to a teammate on the Sales roster. The control plane fires, queues for a person or refuses every action, and logs it.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room with SERA leading it. Every Nagent agent is a configuration of ten declared parts, not a program. A named engineer works with you through onboarding and a named operator through delivery.

SERA reads the CRM digest, named deals, the team record and the team roster. Gmail or Outlook and Google Calendar connect for e-mail and meetings. Other outside systems are reached through connectors granted one key at a time, with every tool action ticked on individually before any agent may use it.

LIPI writes in the workspace's voice and from its offerings and knowledge only, and the Knowledge Hub every agent reads holds the brand, the buyers, the rules and the results. Each e-mail stays a draft until a person approves it, so a rep can make it specific before it is sent.

Yes. People and agents share one team room, and everything said in it is attributed and kept. The agents draft and report; your people approve sends, take meetings and own the customer relationship. A booked meeting is handed to a named person, who owes an acknowledgement before it escalates at 36 hours.

Once the accounts are in the CRM. Choosing Sales at sign-up installs the team in one room, deals reach it once they are in the CRM, and with no pipeline in the record SERA names who can import it. Every agent starts with each action waiting for a person and earns more room on its record.

On Nagent, the AI agent for market research is DRIS. He samples how ChatGPT, Claude, Gemini, Perplexity and Grok answer your buyers' questions, records the citations, and tracks ranked keywords, Search Console, traffic and competitors. What he finds becomes a ranked queue of gaps, each with the reason, the fix and the confidence attached.

A tool waits for a query. DRIS works toward an objective a person sets, measures the same targets the same way before and after, and proposes fixes with their reason and confidence. He works in the Marketing team room beside MIRA, NIA, HOOK and Echo, so the people who own the number see the work in one shared thread.

You do not build DRIS: he arrives on the Marketing team already knowing the job. Every Nagent agent is built from ten declared parts, so changing one changes exactly one thing. Search Console and Google Analytics 4 connect through connectors granted one key at a time, and a named engineer goes with you through onboarding.

Every DRIS figure carries the run count, the window and the engine it was measured on, and the exact prompt behind each result is recorded alongside the citation. The prompt set is fixed per account, so a trend means something. Facts about your company live in the Knowledge Hub, the single source of truth every agent reads.

Yes. DRIS measures each target against the baseline it started from, so earlier runs stay the reference point for later ones. The platform keeps one memory every agent reads before it works: the Knowledge Hub for the brand, the buyers, the rules and the results, with each agent's own memory on top keeping what worked and what failed.

Nagent does not promise a set time. DRIS comes with the Marketing team, which installs as one unit: a roster, a lead and a room. A person sets his objective and baseline and connects Search Console and Google Analytics 4. A named engineer goes with you through onboarding, and a named operator through delivery.

DRIS covers answer engine sampling and citations, gaps, the site audit, ranked keywords, Search Console, traffic and competitor tracking, with reports that say what moved. For paid creative research, HOOK's SCOUT finds competitor and category ads worth studying, and MINE mines the swipe library for what is working and why. Each runs under its own approval rules and budget.

At Nagent, an AI agent for ecommerce product description writing is CREA, the Content team's lead. She drafts product copy from your Products and services, Knowledge Hub and confirmed brand voice, removes any sentence she cannot source, and returns the work as a file for a person to review and approve. Images go to RUPA.

A general chat assistant fills gaps with confident generalities. CREA writes from what your workspace holds, so product names come from Products and services and every claim traces to a source. She works in a team room your people share, hands images to RUPA, and every action is logged and held for approval until trust is earned.

Yes. The voice comes from your brand book and stays a draft until a person confirms it. The same house style rules are handed to the writer and to the reviewer, so every description is written and checked against one set of rules, and decisions recorded in the room can be cited back on the next brief.

A plan that includes the Content team, which gives you CREA and the Content team room. Then fill in Products and services, your brand book and the Knowledge Hub, including product documents connected through Google Drive. For product images, a product shot in the Knowledge Hub or the Asset Library lets RUPA place the real product.

The SEO audit in Nagent today runs on blog posts, checking the title, meta description and readability, with internal links only to entities and posts the site already has. For organic visibility across your site, DRIS scores pages on six dimensions and hands content gaps to the Content board as tasks, at most ten a day.

The draft is written without invented detail. A sentence with no source to cite comes out, and CREA says what source would let it back in, so a thin product record produces a shorter description and a clear request for the missing information rather than made-up features or specifications.

The Content team's listed integrations are Google Drive, whose documents are read into the Knowledge Hub, and a Beta site connection that lets your site read published posts from the workspace. No storefront platform is among them today, so a person moves approved copy into the store. Outside systems are reached through connectors granted one key at a time.

The AI agent for SEO content on Nagent is DRIS working with CREA's Content team. DRIS audits the site, reads ranked keywords and Search Console, and ranks the content gaps. CREA's team takes each gap as a task, then researches, drafts, edits, reviews, illustrates and links a blog post grounded in your own sources, and waits for a person to approve it.

A general writer fills gaps with confident generalities. CREA's team drafts from your brand profile, products and Knowledge Hub passages, may assert only what those support, and appends a Sources section. The same house style goes to the writer and the reviewer, and an independent fact and compliance check runs at publish and holds the post if it finds a blocker.

Not to write or review. A person starts a run in Blog Studio by typing a topic and approves the post in the same place. Search Console, Google Analytics 4 and Google Drive connect as integrations. Your own site reads published posts through the site connection, in Beta, with its own site key. A named engineer goes with you through onboarding.

The voice comes from your brand book and goes to both the writer and the reviewer, so a rule the review enforces is a rule the writer was told. A voice extracted from your material stays a draft until a person confirms it. Products are named from Products and services, so product names are never invented.

Each Blog Studio run writes one post from one topic, started by a person on that surface. A daily blog budget applies: when it is spent, new runs stop until the next day, and the Content Studio shows what the day has spent. DRIS hands the Content team at most ten content gaps a day.

DRIS scores every target on six dimensions: coverage, share of voice, entity strength, recommendation position, sentiment and whether the page is legible to an agent. He turns each gap into one ranked Queue with the reason, the fix and the confidence attached. Content gaps reach CREA's board as tasks, and a person decides what runs.

DRIS keeps measuring. He reads Search Console and Google Analytics 4 daily and re-scores the same targets on the same six dimensions, so movement shows against the baseline he started from. His reports say what moved, with the run count, the window and the engine on every figure.

An AI agent for newsletter writing at Nagent is CREA, the Content team's lead, working in a room your people share. Unlike a general chat assistant, she drafts from your Knowledge Hub, products and confirmed brand voice, removes any sentence she cannot source, and returns the edition as a file to review. A person sends it.

A plan that includes the Content team, which gives you CREA and the Content team room. Then fill in what she writes from: your brand book for the voice, Products and services for what you sell, and the Knowledge Hub, including documents connected through Google Drive. The more a team puts in, the better every draft gets.

Yes. The voice comes from your brand book and stays a draft until a person confirms it. The same style rules are handed to the writer and to the reviewer, so a rule the review enforces is a rule the writer was told. Decisions recorded in the room are permanent and the team's agents can cite them back.

No. Nothing is published, scheduled or sent by CREA, and sending an email is never started from a team room, only where it can be approved. A separate email creator for brand-voiced e-mail templates is in Beta in the Email templates editor, and the Content install does not seat it yet.

What the workspace has shared: Knowledge Hub passages, including documents read in from Google Drive, Products and services, the brand profile and the brand book. CREA holds every factual claim to those sources. If nothing supports a sentence, it comes out, and she tells you which source would let it back in.

No. CREA comes with the Content team, already set up as its lead, and you brief her in plain words in the Content team room. Every Nagent agent is built from ten declared parts, so it is a configuration rather than a program, and changing one part changes exactly one thing.

Every factual claim must trace to what the workspace shared, or it comes out. The draft returns as a file for a person to review, and any action that needs approval waits as a card in the thread or the Inbox. Under the stricter governance profile, content always asks, and sending stays with a person.

On Nagent, an AI agent for real-estate lead qualification works from the record. SERA reports where each lead stands from the CRM and says what is missing, LIPI drafts follow-up e-mail in your voice for a person to approve, and a booked meeting becomes a handoff card owned by a named person.

Choosing Sales at sign-up installs the Sales team into the company's workspace, seated in one team room with SERA leading it. Getting leads into the CRM is the step that matters: if the record holds no pipeline, SERA says so and names who can import it. A named engineer works with you through onboarding.

A lead that arrives at night is in the CRM when your team starts. Drafting runs on its own, but sending waits: by default, nurture steps are drafted for a rep to review, and sending is never started from a room. So no lead gets an unreviewed e-mail at any hour, and a person decides what goes out.

A person. SERA reports a deal's stage, owner, value and last activity from the record, says what is missing where it is silent, and never invents a contact, a value or a date. Your team judges readiness against the criteria recorded in the team record, and SERA hands the follow-up to whoever owns it.

An outside CRM can be connected as a tool, under a connection only your workspace can use, through a connector granted one key at a time. Every tool action is ticked on individually before any agent may use it, and SERA answers pipeline questions from the CRM digest in the team record.

They stay in the CRM. LIPI, in Beta, can draft a nurture sequence step in the workspace's voice, and nurture steps wait for a rep to review unless the workspace changes that setting. When the lead comes back, the record shows where it stands and SERA can report it.

A chatbot follows a script. Nagent's Sales team works beside your people in one room, from one record: SERA answers from the CRM, hands work to the teammate who owns it and names what is missing instead of guessing. Every action is fired, queued for a person or refused, and written to the action log.

The AI agent for ad video creation on Nagent is Alpha. She plans a video shot by shot from a brief, quotes it in credits and waits for a person to approve the storyboard. Then she casts, renders, cuts and proofs it. The words come from HOOK's brief or your copy, and HOOK's GATE checks paid cuts before they ship.

No. Describe the video in Alpha's studio or pick a preset. From another team's room, writing @ALPHA with a request puts a brief on the Content board for CREA with Alpha named on it. Once assets, budget and the brief are in place, the preset carries each job from brief to proof by itself.

A product shot only ever renders from your product's own image in the RUPA library, never from words alone, and PROOF compares the frames against it for pack, label and colour. Your workspace logo is placed in the template's logo slot, and every word on screen comes from the brief or copy a person supplies.

Yes, within a preset. Each preset renders natively in the ratios it offers, such as 9:16 and 4:5 for a Meta Reels ad or 16:9 and 9:16 for a YouTube pre-roll, and each ratio is cut separately. Variants of a delivered cut, including a resize, captions or a new language, reuse the shots already rendered.

Alpha plans and quotes first, and nothing renders until a person approves the storyboard. The words are never hers, product shots start from a real product image, every synthetic presenter carries a disclosure layer, and PROOF checks each rendition before release. A variant renders again only what changes, so nothing is paid for twice.

Alpha does not distribute. She never publishes, traffics or touches an ad account. Paid cuts go back to HOOK's GATE, and PACK assembles a pack for a person to upload by hand. HOOK reads a connected Meta account, or an export from any ad platform uploaded as a file, so PULSE can rank the ads.

Nagent quotes no set deployment time. Alpha arrives with the Content team, and a named engineer goes with you through onboarding. Per job, a render takes 30 minutes to 2 hours of wall-clock, and the job carries a clock that pauses while it waits on a person, an input or a cap.

A bot fires messages from a script. On Nagent the closest thing to an AI agent for LinkedIn prospecting is NORA, and she does not message on LinkedIn at all. She researches prospects that fit your customer profile into sourced dossiers and runs email outreach, while any social message is written as a template and sent by a person.

NORA does not send connection requests or messages on LinkedIn or any other social network. She sends email. Templates for social networks can be written, and a person on your team sends them, so whatever goes out on LinkedIn is sent by a person rather than by an agent.

NORA comes with the Sales team. Choosing Sales at sign-up installs the team into your workspace, seated in one team room, and a named engineer works with you through onboarding. Prospecting starts from the customer profile you save, and on a new workspace every send waits for a person.

From research. Each prospect's dossier carries its sources, and drafts are written from that dossier and from your workspace's own knowledge and offerings, so every claim about the reader traces to something she found. VANI writes the emails in your workspace's voice, and a person reads each draft before it is sent.

NORA qualifies the accounts she finds against the customer profile you saved before she researches them. After outreach, she classifies each reply as interested, objection, scheduling or unsubscribe. A qualified reply becomes a meeting and a handoff card on her Meetings board, which a named person acknowledges.

Nagent keeps the record. Each prospect's research is kept as a dossier with sources, every action NORA takes is written to the action log, and everything said in the Sales team room is attributed and kept. When a reply arrives, NORA classifies it so the cadence stops or changes course.

Research and contact reveals run against a daily outreach budget and a cost-cap exception queue rather than open-ended. Saved campaigns carry a cap on profiles per run, and every agent on Nagent has a daily and a monthly spending cap, so a longer list never means an open-ended bill.

At Nagent, an AI agent for FMCG content creation is CREA's Content team. CREA drafts briefs, articles, e-mails, posts and scripts from your products and sources, and RUPA, her image lead, renders product photoshoots, social posts and ad creatives with your brand guidelines attached. Nothing is published until a person approves it.

It is a team, not a single prompt. CREA, RUPA and your people share one room, its thread and its pinned documents. Every draft is grounded in your products, Knowledge Hub and confirmed brand voice, every image carries your brand guidelines through Brand Lock, and every action is logged and held for approval until the agent earns trust.

Brief by brief. CREA drafts what each brief asks for, from the products and passages the workspace holds, so name the market and its rules in the brief. For visuals, a RUPA reference image can carry the localise it intent, and each of the five reference intents says what it keeps and what it drops.

No. The Content team arrives already structured, with CREA leading and RUPA making images. You brief CREA in plain words in the Content team room and RUPA in her studio, where nineteen presets in seven groups fix the size and plate count. Every agent is a configuration of declared parts, not a program.

It keeps claims sourced and a person on the last step. CREA removes any factual claim she cannot trace to your documents, and blog posts pass an independent check of facts and compliance that fails closed. RUPA's Brand check reads briefs against your banned words and flags what it cannot see. A person approves before anything goes live.

CREA writes briefs, articles, e-mails, posts and scripts, and the team produces illustrated blog posts. RUPA renders packshots, lifestyle scenes, carousel frames, social posts for Instagram, LinkedIn, X, Facebook and Pinterest, and ad creatives for Meta, Story and Reel covers and LinkedIn. FOLIO turns your documents into long-form PDF and DOCX publications.

A plan that includes the Content team, which gives you CREA, RUPA and the Content team room. Then add your brand book with its voice and banned words, your logo, Products and services, and product shots in the Knowledge Hub or the Asset Library. Documents connected through Google Drive are read into the Knowledge Hub.

As an AI agent for ad video creation for FMCG companies, Nagent splits the work. HOOK writes angles, hooks and scripts from your brand pack and checks every claim. Alpha plans, quotes, renders and cuts the video from a RUPA keyframe of the real pack. A person approves the storyboard, reads the proof and spends in the ad platform.

Nagent quotes no set time. Before the first job, HOOK needs a complete brand pack, and Alpha needs a product keyframe from RUPA, the copy or HOOK's brief, and a budget. A named engineer goes with you through onboarding. Each render then takes 30 minutes to 2 hours of wall-clock.

Each product shot starts from that product's own keyframe in the RUPA library, so the right pack, label and colour go into each video, and PROOF compares the frames with that image. Voice, claims and banned language come from the workspace's brand pack, which every HOOK brief reads before it is written.

It works alongside them. HOOK is built for the paid social manager and the creative strategist, not instead of them. A person pins the hooks before SCRIPT writes full scripts, approves each storyboard and reads every PROOF report, and a person uploads and spends in the ad platform.

GATE checks every claim against the brand pack, which lists the claims the brand may make with their proof and its banned language, and holds what the pack does not support. Overriding a block needs a second person. Alpha's words come only from the brief, and every synthetic presenter carries a disclosure layer that PROOF enforces.

Alpha's product ad presets are a Meta Reels ad, a Meta feed video, a YouTube pre-roll and a LinkedIn video ad, each rendered natively in its ratios, such as 9:16, 4:5, 1:1 and 16:9. HOOK's channel packs today are Meta cold 9:16, Meta retargeting 1:1, TikTok Spark and Google RSA. Alpha never traffics an ad.

A standard tool turns one prompt into one clip. Here HOOK checks every claim against your proof before a script ships, Alpha quotes before rendering and renders the product only from its real image, PROOF checks each rendition, and VAULT keeps what each campaign taught for the next brief.