Team module - Overview

The Team module (sidebar → Team, route #/team) is a chief-of-staff and organization-design desk. The agent designs company and team structures from strategy, writes role sheets and RACI matrices, builds hiring plans and interview kits, cascades OKRs, prepares ops reviews - and can spawn a virtual team of AI agents, each with a role and a deliverable, to execute a mission in parallel.

The interactive organization

The top of the view shows your org chart: members with an avatar, a name, a role, a specialty, and their key skills, laid out by reporting line (top level first, then each manager's reports below). It ships with a starter organization - Atlas (Chief of Staff) and six leads: Forge (Engineering), Nova (Marketing), Vector (SEO), Compass (Sales), Mentor (HR & Talent), Scribe (Documents).

• Call on a member card opens the chat and routes the conversation to that member: their workflow command (/forge, /campaign, /seo, /leads, /studio, /team…) plus their role, specialty, and skills as the persona.
• Add member / edit / remove: build your own organization - name, role, specialty, emoji avatar, workflow command, skills, and who they report to. Stored server-side (~/.navin/webui/team-roster.json), shared by every chat.

Calling members from any chat

Type @ in any composer: your team members appear first in the mention palette (avatar + role). Picking one rewrites the message so it runs as that member - e.g. @nova launch a campaign for our new plan becomes /campaign You are Nova, Marketing Lead… launch a campaign for our new plan. Members, CLI apps, and MCP servers share the same @ palette.

How it works

1. Open Team in the sidebar.
2. (Optional) Type a brief at the top: your company, stage, headcount, goal. It is attached to every action.
3. Pick an action card in one of the four groups - Design, Staff, Run, Grow (see Actions).
4. The chat opens and /team is sent automatically with the action's specification and your brief.
5. The agent works and saves every artifact as files under org/ or team/ in the workspace.

Direct usage in any chat:

/team design the organization for a 12-person SaaS startup moving from founder-led sales
/team spawn a virtual team to produce our investor data room: analyst, writer, reviewer

The /team command

The virtual AI team

The distinctive capability: the agent can staff a mission with real subagents, not just documents.

1. It derives 2-5 roles from the mission, each with exactly one deliverable and an exact output path.
2. It writes a complete brief per agent (mission, context, skills to apply, boundaries) - subagents do not share the parent's context.
3. It spawns them with the subagent tool, in parallel or as a pipeline (research → write → review).
4. It tracks the roster in team/roster.md, verifies every deliverable against the brief, and merges the results itself.

Orchestration patterns available: parallel specialists, pipeline, producer + critic, manager + squad. Teams are capped at 5 concurrent agents; beyond that the agent asks you first.

Design principles the agent applies

• Structure follows strategy - the org is derived from outcome streams and workload, never copied from a template.
• One owner per outcome - no role without owned outcomes and KPIs; no outcome with two owners.
• Minimal viable org - the smallest structure that works, plus a growth path with metric-based hiring triggers.
• Explicit decision rights - RACI per key process, escalation paths, no orphan decisions.

Continuous operation

Combine with Navin's autonomy features:

• /goal keep the OKR tracker up to date and flag at-risk key results weekly - a sustained goal.
• Cron jobs for recurring ops review packs and team health pulses.
• /pilot to route heavy design work and quick updates to different models.

Tips

• Give real context in the brief: stage, headcount, revenue, what's breaking - the org design gets dramatically better.
• Chain the groups in one chat: design → role sheets → hiring plan → interview kits.
• For missions, prefer the virtual AI team card when there are ≥3 separable workstreams; solo execution is faster below that.