Setup and daily use are in the Usage Guide.
TeamAI keeps a team's skills, rules, agents, learnings, models, MCP, and other AI assets in a git repo, and syncs them into each person's AI tools.
| Layer | What it does | Read |
|---|---|---|
| Team Execution | The same skills, rules, and hooks on every machine | Core Harness, Distribution mechanism |
| Team Context (beta) | Agents can search what the team learned and how the code is built | Team Context |
| Team Improvement (beta) | Sessions feed lessons back to the team, and you see how the team uses AI | Dashboard, Usage reporting |
What each agent supports is in the README's Product Overview. Git hosts: GitHub, GitLab, GitCode, CNB, TGit, and any private Git service.
teamai push → branch + PR → review and merge
↓
SessionStart hook → teamai pull → each AI tool
| Asset | Location |
|---|---|
| Skills | skills/<name>/SKILL.md |
| Rules | rules/*.md |
| Shared instructions | claudemd/*.md |
| Agents | agents/<name>.yaml |
| Team culture | culture.md |
| Env | env/env.yaml |
| Hooks | hooks/hooks.yaml |
| MCP | mcp/mcp.yaml |
| Models | models/models.yaml |
| Packages | packages in teamai.yaml |
| Docs | docs/ |
| Learnings | learnings/ |
| Codebase knowledge graph | teamwiki/ |
An asset under a <namespace>/ subdirectory goes only to the roles and projects that declare it. See Distribution mechanism.
teamai recall enable # off by default
teamai recall "port conflict"
teamai import --from-repo https://github.com/org/repo # codebase knowledge graphWith recall on, an agent searches the team's lessons before a task. When a session has repeated corrections or retries, its end suggests sharing the lesson with the team.
teamai digest # weekly team digest
teamai session save # save a redacted session summary
teamai dashboard # local dashboard
teamai recall maintenance --prune --dry-run # find lessons nobody uses