Slack’s hosted MCP server at mcp.slack.com/mcp exposes search, channel and thread history, messaging, user profiles, and canvas actions over streamable HTTP. Connect with a token from a Slack app (for example a bot user OAuth token with the scopes your tools need), then drive day-to-day team workflows from an MCP client without building a custom Slack integration.
https://mcp.slack.com/mcp
Claude Sonnet 4.5
MCP Playground runs 30+ models on the same workflow: switch anytime, or use Compare mode to run several in parallel and balance quality vs. cost.
Slack app installed to the workspace. Use the bot user OAuth token (`xoxb-…`) and scopes for search, history, and chat:write (see Slack MCP docs for the full set).
How models use it and what it is built for.
The Slack MCP server is built for LLM use: you get tools for content search (messages, files, channels, people), reading a channel or thread history, sending or drafting messages, creating or updating Slack canvases, and reading user profiles. Workspace admins can approve the calling app, and the same product rate limits apply as the Web API. For Agent Studio, configure the pre-filled URL, install your app to the workspace, and pass the token your client documents — typically a bot `xoxb-` token with chat, search, and history scopes.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
What did I miss in #engineering since yesterday? Summarise the main threads.
Search for every message mentioning "SLO" in the last 14 days and group by channel.
Who is in the #releases channel and when was the last post?
Draft a short standup I can post to #standup based on my recent thread activity.
This is not a single-model product: you get the same MCP connection with 30+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Slack
Claude Sonnet 4.5
Claude Sonnet 4.5 is stronger on long thread synthesis and safe messaging. The Slack template uses Haiku 4.5 for lighter search; compare in Agent Studio.
Open MCP Agent Studio with the connection pre-filled. Add your token, pick any of 30+ models, and start chatting — no install required.
Try Slack in Agent StudioCommon questions about connecting, scoping and using it safely.
It is Slack’s hosted Model Context Protocol service at mcp.slack.com/mcp. It exposes your workspace in tools so an AI can search, read conversations, post messages, and work with canvases, subject to app approval, scopes, and the same rate limits as Slack’s APIs.
Create a Slack app, install it to your workspace, and grant the scopes the MCP tools require: typically search, channel history, and chat:write for posting. Copy the bot user OAuth token and supply it in your MCP client. Admins may need to approve the app for the workspace.
The app only sees channels and DMs the token can access. Start with a narrow set of test channels, avoid broad chat:write until you are comfortable, and review messages the model proposes before they go out if your client does not require approval for every post.
Slack partners with several clients, and any MCP client that can send streamable HTTP to the endpoint with a valid token can work. In this project, use MCP Agent Studio: choose the Slack template or paste mcp.slack.com/mcp and your token.
Claude Sonnet 4.5 handles long thread synthesis and disambiguation well. For cheap high-volume search-only tasks, try Claude Haiku 4.5. Compare them side by side in Agent Studio on a real catch-up question.