Slack MCP Server

Archived reference MCP server that lets Claude and other MCP clients read and post in a Slack workspace via a bot token.

Free· Free / open-source (MIT). Requires a Slack workspace and a Slack bot token; Slack itself is free/paid.MCP Servers

Visit website →

Best for

Developers prototyping an MCP-based Slack assistant, or teams that want a small, auditable reference server to fork and extend into their own internal Slack integration.

Skip if

Production teams that need ongoing maintenance, DM/private-channel/search coverage, or enterprise-grade audit and compliance — use an actively maintained community fork or a first-party Slack app instead.

Slack MCP Server is a reference Model Context Protocol server originally published in the official modelcontextprotocol/servers repository that lets an MCP-capable client (Claude Desktop, Cursor, Zed, Continue, custom agents) read and write to a Slack workspace via a bot token. It exposes eight focused tools: slack_list_channels, slack_post_message, slack_reply_to_thread, slack_add_reaction, slack_get_channel_history, slack_get_thread_replies, slack_get_users, and slack_get_user_profile. Together those cover the most common assistant workflows: browsing public channels, posting messages and thread replies on behalf of a bot user, adding emoji reactions, pulling recent history for summarisation, and looking up members and their profiles for @-mentions and routing. The server is a thin, auditable TypeScript wrapper around Slack's Web API and is packaged for both npx and Docker; setup is a matter of creating a Slack app, granting the channels/chat/reactions/users OAuth scopes, and passing SLACK_BOT_TOKEN plus SLACK_TEAM_ID as environment variables. It is aimed at developers and platform engineers who want an LLM assistant to triage channels, draft or send Slack updates, summarise long threads into a digest, or fan a decision out as a reply, without writing bespoke Slack integration code. Note the important status caveat: the original repo was archived by Anthropic on 29 May 2025 and is no longer maintained; the code still works but new Slack API changes, scope updates, and security fixes will not land upstream. Several community forks (notably korotovsky/slack-mcp-server and zencoderai/slack-mcp-server) have picked up active development and add features like DM/private-channel support, search, and stealth-mode auth — most teams standing up a new integration today should evaluate one of those forks and use this archived server mainly as a reference implementation of the MCP tool shape.

Editor's take

A clean, minimal reference for how an MCP server should look, and still the fastest way to give Claude Desktop hands on a Slack workspace in five minutes. But the archived badge matters: I would use this to learn the pattern, then move to korotovsky/slack-mcp-server or a hand-rolled internal server for anything a real team will depend on.

— The AI Tool Bible editorial team

Pros

✅ Official reference implementation from the modelcontextprotocol project — small, readable TypeScript that documents the MCP tool pattern well
✅ Covers the eight highest-value Slack primitives (list, post, reply, react, history, thread, users, profile) with minimal ceremony
✅ Ships as both an npx package and a Docker image, so it drops straight into Claude Desktop, Cursor, or any MCP client with a JSON config snippet
✅ Standard Slack bot-token auth (xoxb-) with clearly scoped OAuth permissions — easy to reason about and to revoke
✅ Open-source under MIT, so forking or vendoring for internal hardening is straightforward
✅ Good starting point for learning how to write your own MCP server against a REST API

Cons

⚠️ Archived by the maintainer on 29 May 2025 — no upstream bug fixes, security patches, or Slack API compatibility updates
⚠️ Read-only surface for channels (public channels only by default) and no DM, private-channel, search, files, or canvas support out of the box
⚠️ No pagination helpers or rate-limit backoff beyond what the Slack SDK provides, so bulk history pulls in large workspaces can be fragile
⚠️ Requires a workspace admin to install a bot app and mint a token, which is a real blocker in locked-down enterprise Slack tenants
⚠️ Bot-token model means every action is attributed to the bot user, not the human operating the assistant, which complicates audit trails

Use cases

Channel triage and summarisationAutomated thread digestsPosting AI-generated status updatesCross-channel search and reportingStandup and retro note postingOn-call escalation repliesUser and profile lookup for routingReaction-based workflow signallingMCP server reference implementation