How to connect Claude Code to your team chat with MCP
Give Claude Code, Codex or Cursor read and post access to your team channels, docs and tasks over MCP.
What this gets you
Your coding agent usually knows your repo and nothing else. The decisions, bug reports and context live in team chat. Connecting the agent to chat over MCP, the Model Context Protocol, lets it read a channel, search past messages, open a doc or create a task without you copy-pasting.
Dock runs an MCP server on your own workspace. You create a personal access token, register the server once in your agent, and the tools appear.
Tools the agent gets
Read
List channels and chats, read messages, search messages, search and read docs, read meeting transcripts.
Write
Post and send messages and add reactions, in the places your token can reach.
Work
List task lists and tasks, create a task, mark one complete.
Set it up in three steps
1. Create a token
In Dock open the Connect AI page and create a personal access token. It acts as you, so treat it like a password.
2. Add the server to your agent
For Claude Code run `claude mcp add --transport http dock <your-dock-url>/mcp --header "Authorization: Bearer <token>"`. The Connect AI page prints this line with your own URL and token filled in.
3. Check it and ask
Start `claude`, type `/mcp` and confirm dock is connected. Then try: "List my Dock channels and summarise what happened today."
Other agents
The same server works with Codex CLI (a `[mcp_servers.dock]` block in `~/.codex/config.toml` with the token in an environment variable), Cursor (`mcp.json`), Windsurf and Gemini CLI. The Connect AI page has a copy-paste snippet for each.
Tip: give an agent its own bot account instead of your personal token when it will post on its own. Bots only see the task lists and channels you grant them.
Good first prompts
Catch-up: "Read #engineering since yesterday and list decisions and open questions."
Triage: "Search messages for 'login error' this week and create a task for each distinct bug."
Handoff: "Read the launch doc and post a three-line status to #launch."
Safety notes
A token carries your permissions, so keep it out of repos and shared shells. Revoke it from the Connect AI page when you stop using it. Ask the agent to draft before it posts in busy channels, and prefer a dedicated bot for anything unattended.
This is a young area: agents can misread context and tools change quickly. Treat output as a draft you review.
Questions
What is MCP?
The Model Context Protocol is an open standard that lets AI agents call tools and read data from other apps through a common interface.
Does this send my chat to an AI provider?
Only what the agent reads is sent to the model provider you already use with that agent. Dock itself does not forward your chat to an AI vendor through this connection.
Can the agent see private channels?
It sees what your token's account can see. Use a bot with granted access to limit it.
Do I need to self-host?
Dock is self-hosted, so the MCP server runs on your own instance.
Connect your agent to Dock
Free to self-host. Includes the Connect AI page and MCP server.
Download for Mac