pi-teams

Microsoft Teams integration for the pi coding agent.

pi acts as you: it reads what you can read, and anything it sends comes from your account, in your name, in the right thread. Multiple Teams accounts and multiple company tenants live side by side, and a configuration file decides exactly which teams, channels, chats and people pi may touch.

Installation

# From npm
pi install npm:@patimweb/pi-teams

# From a local checkout during development
pi install /path/to/pi-teams

Quick start

1. Register an app in Entra ID — once per organization. Two settings decide whether sign-in works at all (a redirect URI and public client flows), so follow docs/entra-app-registration.md rather than guessing from the portal.

2. Configure the account:

   teams_setup:
     name: work
     tenantId: contoso.onmicrosoft.com
     clientId: <application (client) ID>
     setDefault: true

3. Sign in: run /teams-login. Your browser opens at the Microsoft sign-in page and closes itself when you are done — nothing to type.

4. Check it works: /teams-status, then /teams-inbox.

If anything goes wrong, run teams_doctor before anything else: it checks the configuration, the sign-in, which scopes were actually granted, and whether live calls succeed, for every configured account.

What it can do

Read and write chats and channel threads, search your Teams history, react to and edit your own messages, download the files and pictures a message carries, see who is available, and schedule, change or cancel meetings — across several accounts and tenants at once. The full list is in docs/tools.md.

Listen mode is the one feature that acts without being asked: pi polls your chats and answers incoming messages as you. It is off by default, never starts on its own, and is worth reading about before switching on — docs/listen-mode.md.

Documentation

| Document | What it answers | |----------|-----------------| | entra-app-registration.md | How do I register the app, and who has to approve which permission? | | sign-in.md | How does pi sign in, where is the session kept, what happens on a headless machine? | | configuration.md | Every key in pi-teams.json: accounts, tenants, safety levels, scope rules, the AI footer. | | tools.md | Every tool, command and prompt — and how to address a chat, a channel or a mention. | | listen-mode.md | How pi answers on its own, and how to keep that narrow. | | security.md | What pi may do in your name, what stops it, and what it writes to disk. | | troubleshooting.md | A symptom-to-cause table, including the AADSTS codes. | | architecture.md | Module map and development setup, for contributing. |

The bundled skill (skills/teams-collaboration/SKILL.md) is documentation too, but for the model rather than for you: it is what tells pi how to write a Teams message, when to read before replying, and which tool fits a request.

A note on what this is

pi sends messages that are indistinguishable from ones you typed. There is no "sent by a bot" marker in Teams, which is why an AI disclosure footer is appended by default and why the default safety level asks before every write. Both are configurable; both are deliberate. See docs/security.md before you turn either off.

License

MIT