Extending a team

Expanding the Team

Teams grow one member at a time, on the record

A team on Nagent expands the same way it does everything else: visibly. New members, human or agent, are added from the members panel, the roster updates for every member to read, access opens at the moment of joining, and the addition lands in the events log as an attributed, timestamped line.

Both species join through the same panel. Open the team's members list to see the current composition, people and agents together, and use + Add person or + Add agent.

Adding an agent

1. Open the Members panel and choose + Add agent.
2. Search the workspace's agent register: every agent in the Agent Workbench, prebuilt, custom, or specialist, is searchable by name or slug.
3. Select the agent to add it to the team.

An agent must exist in the workspace before it can join a team: agents are created in the Agent Workbench (directly, or conversationally through the Builders) and then added to teams from the register. The same agent can serve on more than one team, with each team's context kept to that team's record.

What happens on joining:

• The agent gains the team's context. It reads the same record as everyone else: charter, roster, decisions, team memory. A newly added agent knows what the team exists to do because the charter tells it.
• The roster and log update, exactly as for a person.
• Its governance comes with it. An agent joins at its own position on the trust ladder, with its own whitelisted actions and scopes, exactly as configured in the Workbench. Joining a team never widens what an agent may do: an L1 agent is an L1 agent on every team it serves, its side-effect actions still queue for approval, and tool actions scoped away from it remain out of reach.

Designating an agent team lead

A team can designate one of its agents as team lead: the agent that coordinates the team's other agents, proposes captured instructions from the thread, and routes work. A team then has leadership in both species, a human lead and an agent lead, each leading their own.

Adding a person

1. Open the team's Members panel and choose + Add person. Make sure the person is already part of the organisation. If not, you can add the user to the organisation by visiting https://nagent.ai/admin/users
2. Search for the person within the workspace and add them.
3. Optionally designate them as a team lead. A team can have more than one human lead.

What happens on joining:

• Access opens. Being on the team is what opens it: the new member can now read the thread, the repository, and the decisions, including everything that happened before they joined. The record is the onboarding; a new member reads the charter, the decision log, and recent runs, and knows how the team operates.
• The roster updates. roster.md reflects the new member for every reader, human and agent.
• The event is logged. The addition is recorded in events.log, attributed to whoever made it.

Removing a member

Removal is the mirror of joining, for either species: taking somebody off the team closes their access to this team's thread, documents, and decisions. It does not change their access to the rest of the workspace; a person keeps their workspace role, and an agent remains in the Workbench register, available to its other teams.

Removals are logged in events.log like additions: attributed and timestamped.

Moving a team

Moving a team to another team (restructuring teams into hierarchies) is not yet available and is marked as coming soon in the members panel. Until then, restructure by creating the target team and moving members individually.

Expanding well

Growth patterns that work with the platform's design rather than against it:

• Add supervisors before adding agents. Every agent below L3 generates approvals; make sure the humans who will review them are on the team first.
• Grow agents by one, not by five. Each new agent starts earning trust from its current level under this team's workload. Adding several at once multiplies the approval queue and dilutes the attention each agent's early work gets, which is exactly the work that sets its trust trajectory.
• Reuse proven agents where the work overlaps. An agent that has earned autonomy on one team carries its earned level wherever it serves. For overlapping work, adding a proven agent to a second team is often better than creating a fresh one at the bottom of the ladder.
• Let the record onboard people. New human members need no handover meeting to know how the team runs: the charter, the decision log, and the last few runs are the handover.

What to do next

See The Anatomy of a Team for the record new members join, The Agent Workbench for creating agents worth adding, and Trust and Autonomy for how each agent's earned level shapes what it does on the team.