Skip to main content
Back

Switch Brings Multiple AI Agents Into Slack and Teams Channels

Switch turns Slack, Teams, Discord, Telegram and Mattermost channels into shared rooms where humans and AI agents collaborate on the same thread.

Repo, Full Access
  • Switch is an open-source platform putting AI teammates into Slack, Teams, Discord, Telegram and Mattermost channels.
  • Channels become shared rooms where people and multiple agents work from the same context and history.
  • Agent-agnostic: bring Claude Code, Codex, LangChain or any runtime that speaks the protocol.
  • Fully self-hosted under Apache 2.0 with Commons Clause, so data and context stay on your infrastructure.
  • Installed via a desktop console with builds for macOS, Windows and Linux; setup is guided by an in-app agent.
  • Docs are served over MCP so your own agent can walk you through configuration.

Switch has crossed roughly 3,900 GitHub stars after adding several hundred in the past week. The source-available project turns channels in Slack, Microsoft Teams, Discord, Telegram, and Mattermost into shared workspaces where people and multiple AI agents can use the same conversation history, decisions, and files.

Switch occupies the coordination layer between existing chat services and agents such as Claude Code, Codex, and LangChain applications. Each room has its own members, instructions, tools, and access rules, allowing agents running on laptops or company servers to participate alongside a team.

One room, several agents

Switch maps each messaging channel to a persistent room with a membership list. An agent receives context from the rooms it has joined and can respond according to its assigned role, instructions, and available tools.

  • Agent portability: Teams can connect local coding agents, hosted services, or custom agents that implement the supported protocol.
  • Room-scoped context: Each agent sees the conversations from its assigned rooms, limiting unnecessary context sharing.
  • Tool assignment: Agents can receive access to services such as Google Workspace, Jira, and GitHub.
  • Self-hosted infrastructure: Companies can run the Switch gateway and console on their own systems.
  • Access and cost controls: Administrators can define who may interact with each agent and add guardrails or usage reporting.

Setup through the desktop console

Switch Console provides the primary installation and configuration interface. Builds are available for Apple Silicon and Intel Macs, Windows, and Linux on x64 or ARM64, with AppImage and DEB packages for Linux.

During initial setup, an assistant called switch-buddy creates the first agents and channels, then guides the connection to a supported messaging service. Teams still need to supply and configure the agents, model providers, credentials, and external tools they plan to use.

The documentation is also available through Model Context Protocol, or MCP, a standard that lets compatible AI clients access tools and resources. Claude Code users can add the documentation endpoint with:

claude mcp add switch-docs --transport http https://docs.switchagents.ai/mcp

The setup guide includes equivalent commands for OpenAI Codex CLI and OpenCode. Once connected, those clients can retrieve the Switch documentation while helping configure a deployment.

The channel becomes shared memory

One-to-one agent chats fragment context when several people and agents contribute to the same task. A channel gives the group a durable record of requests, responses, approvals, and handoffs, while its membership model determines which participants can see and act on that record.

The project describes an adoption path that begins with adding a coding agent to a feature channel. More advanced workflows can assign bug reproduction to a triage agent, implementation to a coding agent in a separate room, review to a person, and test deployment to an operations agent. Room membership and channel history remain the coordination primitives throughout the workflow.

Costs and constraints

Area What developers should know
Project maturity The repository had 121 open issues. Teams should test upgrades and integrations before production use.
License The repository combines Apache License 2.0 terms with the Commons Clause. That combination is generally described as source-available because the added clause restricts certain commercial uses. Organizations planning to redistribute or sell a related service should review the license carefully.
Data boundaries Self-hosting keeps the Switch gateway under company control. The full data path also includes connected chat services, agents, tools, and model APIs, each with its own storage and retention policies.
Telemetry Server telemetry is opt-in. Reported fields use a fixed vocabulary of enumerated values, numbers, and booleans designed to exclude free-form text.
Operational effort Switch depends on agents and tools supplied by the team. Its value grows with the number of existing workflows that benefit from shared channel context.

Where the model fits

Teams already running coding or operations agents in isolation have the clearest use case. Switch can place those agents in incident rooms, feature channels, and on-call conversations, reducing manual transfers between private agent sessions and group discussions.

Examples described by the project include one incident-response agent charting error rates while another identifies a suspect deployment, a triage agent filing feature requests in Jira or Notion, and an on-call agent drafting a postmortem in the team’s knowledge base.

Frameworks such as LangGraph and CrewAI provide primitives for defining agent behavior and multi-agent workflows. Switch adds a collaboration surface around agents that teams already operate, using familiar channels, membership, and permissions to coordinate work among people and software.