Nagent — Technologies & Systems

Nagent in one sentence

Nagent is multiplayer AI for growth teams.

A company's people and its AI teammates work in one shared context: the same brief, the same record and the same next move, in one workspace. AI teammates hold a role, answer from the company's own material, and ask a person before anything they are not yet trusted to do. They are led by chiefs of staff (MIRA for Marketing, SERA for Sales, CREA for Content), they earn autonomy the way people earn promotions, and they work under a control plane where a person approves anything that goes live.

Every AI teammate in a workspace inherits the same foundations, with no extra wiring:

• Shared context so the team works from one record, not one person's chat history
• Grounding in the Knowledge Hub so answers come from the company's own material, with sources
• Earned autonomy so an agent gets more room only through approved work
• A control plane so every action is proposed, approved, run and audited
• Budgets so an agent stops before it spends past its cap

                         ┌──────────────────────────┐
                         │   PEOPLE ON THE TEAM     │
                         │ brief · approve · decide │
                         └────────────┬─────────────┘
                                      │  team rooms · boards · approvals
         ┌────────────────────────────┼────────────────────────────┐
         │                            │                            │
   ┌─────▼──────┐              ┌──────▼─────┐               ┌──────▼─────┐
   │ MARKETING  │              │   SALES    │               │  CONTENT   │
   │ chief: MIRA│              │ chief: SERA│               │ chief: CREA│
   └─────┬──────┘              └──────┬─────┘               └──────┬─────┘
         │ team leads and bench agents│                            │
         └────────────────────────────┼────────────────────────────┘
                                      │
                ┌─────────────────────▼──────────────────────┐
                │               CONTROL PLANE                │
                │ levels · approvals · budgets · audit trail │
                └─────────────────────┬──────────────────────┘
                                      │
                ┌─────────────────────▼──────────────────────┐
                │  KNOWLEDGE HUB · CRM RECORD · CONNECTED    │
                │  TOOLS · BRAND · CUSTOMER PROFILE          │
                └────────────────────────────────────────────┘

The rest of this document walks those layers from the top down.

How to read this document

• Sections 1 to 3 set the frame: the problem, the shape of the product, and a working day on it.
• Sections 4 to 8 cover the systems every AI teammate runs on: departments and teams, team rooms, grounding, the control plane and budgets.
• Section 9 describes the twelve offerings.
• Sections 10 to 12 cover connections, security and how Nagent is packaged.
• Section 13 is a glossary.

1. The problem: everyone uses AI and none of it adds up

Most growth teams already use AI. Each person prompts alone, in a separate tool, and the context lives in that person's chat history. Nobody can say what an agent did last Tuesday or why, the work does not compound, and the number does not move.

The tools that came before fall into three groups:

Nagent starts from a different unit. The unit of work in a company is a team, so the product is built around a team rather than one assistant. A brief, a decision and a next move that everyone can see is worth more than a clever answer one person saw.

2. The shape of Nagent

A workspace is one company. Inside it:

• Departments each own one outcome. Sales owns qualified meetings, Marketing owns reach and pipeline cost, Content owns output and the brand line.
• Chiefs of staff lead each department and report to you: MIRA for Marketing, SERA for Sales, CREA for Content.
• Team leads speak for their teams and route work to the teammate who fits. Bench agents sit under a lead.
• Team rooms are where people and agents talk, brief, decide and hand work back and forth, on the record.
• The Knowledge Hub holds what the company knows, and every answer is grounded in it.
• The control plane decides, before any action, whether it fires, waits for a person or is refused.

Where the plan includes it, a Nagent forward deployed marketer joins the team and owns the number with you.

3. A day on Nagent: a lean team runs its growth number

08:30. The head of growth opens Home. It shows what is waiting on a person: two drafts to approve, one outbound sequence to confirm, a budget alert on an image teammate.

09:00. In the Marketing room she briefs MIRA with an outcome, not a task list: more qualified demand from mid-market SaaS this quarter. MIRA plans it and hands each piece to the teammate who owns it. DRIS takes the answer engine gap analysis, CREA's team takes two articles, HOOK takes paid social angles.

10:15. CREA's team returns a draft researched from the Knowledge Hub, with sources. The content lead sends it back once with a reason, then accepts the second version. Nothing publishes until a person approves.

12:00. In the Sales room, NORA has found accounts that fit the customer profile and researched each into a sourced dossier. The first e-mails wait as drafts for approval.

15:00. DRIS reports how ChatGPT, Claude, Gemini, Perplexity and Grok cite the brand for the tracked prompts, and proposes site fixes, each with its reason. A person approves two and declines one.

17:30. The Actions page shows every action the agents took today, how each was resolved, and what it cost. Nothing went out that a person did not approve or that the agent had not earned the right to do.

4. Departments, teams and the Org chart

Nagent arranges AI teammates the way a company arranges people.

1. A department owns one outcome and is led by a chief of staff.
2. Team leads report to the chief. A lead speaks for its team and routes work to the teammate who fits.
3. Bench agents sit under a lead. HOOK, for example, brings bench agents for creative strategy.
4. A lead can hand work to a teammate on the same roster, one step deep. The teammate's own level decides what it may do, so a handoff never lends the lead's trust to someone else.
5. The Org chart groups agents as Platform agents, Marketing, Sales, Content and Custom agents, chief first, then leads.
6. Platform agents are maintained by Nagent and locked at L3 for every workspace. A workspace's edit to any other prebuilt agent lands on its own copy.

A company picks its departments at sign-up, and each department's team installs into the workspace on first sign-in.

5. Team rooms: talk is work, and it is on the record

A team room is where the people on a team and its AI teammates work together.

1. Type @ to address someone. Addressing an agent asks it to answer; addressing a person records the message.
2. The room reads what kind of message it is: work an agent can do now, an instruction nothing can run yet, or a question.
3. Work an agent can do now starts at once, after the agent states its assumption or asks whether you want the work or just an answer.
4. An instruction nothing can run yet is proposed by the team lead as a card. Only a person can confirm it.
5. A confirmed task lands on the board, attributed to the person, and the agent named on it picks it up and says so in the thread.
6. A question gets an answer in a thread, grounded on the team's record, with sources to open.
7. A result comes back as a file. Accept puts it on the record; Send back returns it with a reason.

Each room keeps its own record: a charter, a roster, decisions, tasks, shared documents and the memory the team has agreed to keep.

6. Grounding: agents answer from the company's own material

An agent that cannot cite the workspace's own material says so, because a confident wrong answer costs a team more than an honest miss.

1. Point Nagent at the company website. It reads the pages that say what you sell into the Knowledge Hub.
2. Add more as pasted text, a URL or an uploaded document. Each is indexed so agents can search it.
3. Describe the customer in plain sentences. Nagent works out the targeting and shows what you wrote apart from what it concluded.
4. Confirm offerings. A drafted offering reaches no agent until a person makes it active.
5. In a team room, a reply is built from that team's record and carries its sources.
6. If nothing relevant is found, the agent says it cannot cite an answer and asks for the source.
7. If a reply states a figure it cannot cite, the reply is discarded and the honest miss is posted instead.

The brand sits beside the Knowledge Hub: one brand row per workspace that every image, copy block and asset reads from, and a brand voice that applies only after a person confirms it.

7. The control plane: earned autonomy and approvals

Trust is treated the way a company treats a new hire. An agent has a role, a lead and a record, asks before the irreversible step, and earns more room through approved work.

1. Each action an agent can take declares the level it needs to run without a person. Pure reads always run.
2. Agents sit on five levels, from L1 Observer to L5 Principal.
3. When the agent's level meets the action's bar, it fires. When it does not, an approval request is written and nothing runs.
4. A paused or switched-off agent takes no action with a side effect.
5. A governance profile, Standard or BFSI, adds a floor per class of action. Content always asks.
6. Destructive changes ask a person every time, at every level.
7. When a class of action builds a clean record of approved work with no reversals, Nagent offers to let it run with less sign-off. A workspace admin accepts or declines; a reversal resets the record.

Approving is a permission of its own. Each person holds one of six roles, and not every admin or editor can approve.

8. Budgets and credits

Every AI teammate carries a daily and a monthly spending cap.

1. Before a paid call, Nagent checks what the agent has left under both caps, counting money already committed to calls still running.
2. When a cap is reached, the agent's setting decides: queue further paid calls until the window rolls over, pause the agent, or continue and record the breach.
3. A paid tool records its price against the calling agent, exactly as a model call does.
4. Above the agents, a workspace daily limit refuses new paid work; reading, reviewing and approving still work.
5. Expensive jobs, such as a video render, are quoted for a person to approve before they start.
6. Every billable call is recorded on the workspace's credit ledger.

Credits are the unit of agent work on every plan.

9. The offerings

Each offering keeps its own identity. All of them run on the same workspace, team rooms, Knowledge Hub and control plane.

Who it is for

Growth-stage companies whose marketing, sales and content work is carried by a lean team that needs more output and a measured outcome, not another tool: B2B SaaS and technology companies, D2C and consumer brands, and B2B services firms and agencies.

10. Connected tools

A workspace connects the tools its team already uses under its own connection.

1. Nothing is enabled on connect. Each action is ticked individually, labelled if it has side effects, and can be scoped to the agents that need it.
2. When an agent calls an enabled action, the control plane decides: reads run, and actions with side effects run or wait for a person according to the agent's level.
3. Connection health is checked; an expired token turns the set-up card back to "Needs you".
4. Outbound e-mail goes out from the workspace's own sending account and verified domain.
5. Outside services post events to a workspace-specific URL, signed with a secret. Unsigned or wrongly signed requests are refused.
6. A trigger on an agent fires an action on a named event or a schedule, checked against the same rules as if a person had asked.

11. Security, isolation and the audit trail

• ISO/IEC 27001:2022 certified by InterCert.
• SOC 2 Type II programme underway.
• Workspace isolation. Every signed-in request runs inside its workspace. A guard on the data layer adds the active workspace to every read and write that does not state one, and knowledge search refuses to run without the workspace's own filter inside it.
• Audit log. Every admin write is recorded with the user, the action, the record touched and the change. The log refuses updates and deletes.
• Action log. Every agent action is recorded with its trigger, target, the agent's level at the time, cost, duration and resolution: fired, fired after approval, queued for approval, skipped by policy or errored.
• Roles. Each person holds one of six roles, and approving is a permission of its own.
• Private VPC deployment is available on Enterprise.

12. How Nagent is packaged

The live ladder is the pricing page; quote it rather than a remembered figure. As published on 4 October 2026:

• Solo: one seat and a monthly credit allowance.
• Team: ten seats and a larger monthly credit allowance.
• Business: fifty seats, pooled credits and Growth Pods.
• Enterprise: by quote, with private VPC deployment, bring your own model and key, custom guardrails and approval.

Credits per month are the unit of agent work, and annual billing costs less than monthly.

13. Glossary

• AI teammate: an agent with a role, a lead, a level and a record.
• Chief of staff: the agent that leads a department and reports to you. MIRA for Marketing, SERA for Sales, CREA for Content.
• Team lead: the agent that speaks for a team and routes its work.
• Bench agent: an agent that works under a team lead.
• Team room: where a team's people and agents talk, brief and decide on the record.
• Board: a team's confirmed tasks, each attributed to the person who confirmed it.
• Knowledge Hub: the workspace's own indexed material that every answer is grounded in.
• Level: an agent's trust, from L1 Observer to L5 Principal, which decides what it may do without a person.
• Earned autonomy: an agent gaining more room through a clean record of approved work.
• Control plane: the layer that decides whether each action fires, waits for a person or is refused, and records it.
• Credits: the unit of agent work on every plan.
• The number: the growth outcome a workspace is run against.