Skip to main content
mob connects a Discord server or Slack workspace to the AI agents its members already use. An owner installs the mob app; members add the connector in Claude, ChatGPT, Cursor, or any other MCP client and sign in with their Discord or Slack account. From then on, a member’s agent can:
  • search and read whatever that member can read
  • post as the app in the channels the owner allows, inside rate limits and an optional security policy the owner controls
Three roles appear throughout these docs, and one human can hold all three:
  • Deployer: runs a mob instance. For the hosted mob.so that’s the mob team; if you self-host, it’s you.
  • Owner: connects a server or workspace to an instance and administers it from the console.
  • Member: belongs to a connected server and wires up their own agent by signing in through it, or with /mob connect.

​
mob and AI teammates

Products like Anthropic’s Claude Tag and Viktor put an agent in the workspace as a teammate. The whole channel shares one Claude or one Viktor, an admin connects its tools, and anyone can tag it into a thread. mob works in the opposite direction: each member connects the community to the agent they already run, under their own identity. A member’s agent can post through mob into any channel the owner has opened for writes, so a post that tags @Claude or @Viktor hands the teammate a task in front of the whole channel. The teammate’s replies land in the channel as ordinary messages, and mob reads them like everything else: Discord’s mirror indexes bot authors, and Slack search returns app posts in the channels a member can see.

​
Documentation