How to connect Slack to your AI assistant with MCP

TL;DR:

• Connect Slack to Claude, ChatGPT, or Cursor via Composio's managed MCP server, in under 10 minutes.

• Composio's managed MCP server handles over 300M tool calls per month across 1M+ connected accounts: the same execution infrastructure production agent teams rely on.

• The free tier gives you 100,000 tool calls per month with no credit card required, making this a low-risk setup you can test in one sitting.

• Once connected, your AI assistant can read channels, search message history, summarize threads, draft replies, and post updates directly inside Slack.

Every time you copy-paste a Slack thread into Claude to summarize it, you break your focus and burn time. Connecting Slack directly to your AI assistant stops that.

You won't struggle with the initial connection, but keeping it alive is where most setups fail. By using a managed Model Context Protocol (MCP) server, you can give Claude, ChatGPT, or Cursor secure, direct access to your Slack workspace. This guide shows you how to set up that connection in under 10 minutes, manage your permissions safely, and automate your daily communication workflows.

How Slack MCP eliminates manual data handoffs

Think of MCP as a universal connector for AI. The Model Context Layer is the brain. It defines what data your AI assistant can access and what actions it can take inside tools like Slack. The Protocol Layer is the nervous system. It handles how those requests and responses are transmitted between your assistant and the tool.

Without MCP, your AI assistant and Slack operate separately. Every time you want Claude to summarize a thread or draft a channel update, you copy content from Slack, paste it into your AI chat, read the response, and copy it back.

With a Slack MCP setup, your AI assistant calls Slack directly. You type a prompt, the assistant reads the relevant channel, generates a response, and posts it, all without leaving your chat interface. The Composio Slack toolkit exposes Slack's actions as structured, AI-ready tools that any MCP-compatible client can call.

How to connect Slack MCP with ChatGPT Work

Prerequisites

• A ChatGPT account with access to ChatGPT Work (rolling out across Pro, Enterprise, Edu, Plus, and Business plans). We will use the ChatGPT desktop app, which is available on every plan.

• Access to the Slack workspace you want to connect.

• Composio MCP.

Open the MCP settings

ChatGPT Work → Settings → Plugins → MCP

Add the server

MCP → Add server → Name: Composio → Type: Streamable HTTP → URL: https://connect.composio.dev/mcp → Save

Authenticate

Composio MCP → Authenticate → Log in via Composio → Allow access

Start using Slack in ChatGPT

Open ChatGPT Work → Start a new chat → Ask ChatGPT to use Slack → Select the Composio tools when prompted

How to connect Slack MCP with Claude Cowork

Prerequisites

• Claude Desktop with access to Cowork.

• Access to the Slack workspace you want to connect.

• Composio MCP.

Open Customize

Claude Desktop → Customize → Connectors → +

Add the Composio MCP server

Add custom connector → Paste https://connect.composio.dev/mcp → Connect

Authorize in your browser

Connect → Browser opens → Authorize Composio → Continue

Connect your Slack account

Return to Cowork → Ask the agent to connect to Slack or give it a Slack-related task → Authenticate Slack when prompted → Authorize access

For example:

• "Send reminder to marketing channel at 10am"

• "Add reaction to latest team message"

• "Archive inactive project channel after review"

That's it. Composio's tools are now available in Cowork, and your Slack account is ready to use.

How to connect Slack MCP with Cursor

Prerequisites

• Cursor installed on your computer.

• Access to the Slack workspace you want to connect.

• Composio MCP.

Install with one click

Open the Composio Slack + Cursor page → Click "Install in Cursor" → Cursor opens → Add Composio

Or add manually

Open your project → Open .cursor/mcp.json → Add the Composio MCP configuration → Save the file

{
  "servers": {
    "composio": {
      "type": "http",
      "url": "https://connect.composio.dev/mcp"
    }
  }
}

For a global configuration, use ~/.cursor/mcp.json instead of the project-level file.

Authorize

Restart Cursor → Open MCP Tools settings → Find Composio → Click "Connect" → Browser opens → Authorize Composio

Connect your Slack account

Return to Cursor → Ask the agent to connect to Slack or give it a Slack-related task → Authenticate Slack when prompted → Authorize access

For example:

• "Send reminder to marketing channel at 10am"

• "Add reaction to latest team message"

• "Archive inactive project channel after review"

That's it. Composio tools are now available in Cursor, and your Slack account is ready to use.

How Slack MCP automates your workflow

The real friction is not the initial setup. It's in what happens after: whether your AI assistant can reliably execute Slack actions at the moment you need them, at the volume your workflow demands. Composio's managed server routes over 300M tool calls per month across 1M+ connected accounts, and exposes more than 50,000 agent-ready tools, including the full Slack toolkit, through a single MCP endpoint.

That means your AI assistant isn't limited to reading and posting. It can manage channels, trigger reminders, react to messages, search workspace history, and send direct messages, all from one connection. Self-hosted setups give you those same Slack API methods in principle, but carry more setup overhead before your first call lands.

Here's how the two approaches compare:

The self-hosted path is free and runs locally, which suits users who want full control over their data and are comfortable with technical setup. Self-hosted setups require generating OAuth scopes, obtaining Slack authentication tokens, configuring the MCP server, and manually editing your AI assistant's configuration file. For those who want local privacy without the maintenance overhead, Composio offers a Zero Data Retention add-on that prevents data storage while handling all the infrastructure work.

Automate Slack workflows with AI

Before you connect Slack, decide how you'll measure whether it actually saves time. Three things to keep in mind when evaluating your Slack-AI integration:

1. Messages summarized: Track how many times per week you ask your AI to summarize a thread instead of reading it manually. If that number grows, the connection is working.

2. Drafts generated: Count how many Slack messages your AI drafts for you. Even if you edit every draft, that's faster than writing from scratch.

3. Manual handoffs eliminated: Note how many times you would have copy-pasted Slack content into your AI chat before this setup. If that number drops to near zero, the workflow is pulling its weight.

What you can automate once Slack is connected

Once the connection is live, you're not limited to simple message lookups. Here are four workflows that save the most time for solo operators and consultants using Slack daily.

Draft and post Slack messages via AI

Ask your AI to draft a channel update based on context from another tool. After reviewing a project status in Notion or a CRM update in HubSpot, you can prompt Claude to write a summary and post it to your #client-updates channel using SLACK_CHAT_POST_MESSAGE, no copy-pasting required. The Composio Slackbot toolkit extends this further if you want to build a more persistent bot-style setup.

Query historical Slack data with AI

Your AI can search your Slack workspace history to find specific decisions, links, or files buried in old threads. A prompt like "Find the message where we decided on the Q3 deadline in #projects" triggers a SLACK_ASSISTANT_SEARCH_CONTEXT call and returns the relevant result without you scrolling through weeks of history.

Automate Slack thread recaps

With your Slack MCP connection active, you can prompt your AI to fetch a thread via SLACK_FETCH_CONVERSATION_HISTORY and return a structured summary with key decisions, action items, and open questions. For teams using Composio triggers, this recap can run automatically on a schedule or in response to supported Slack events, giving you a summary without needing to prompt it manually.

Use AI to reply to Slack DMs

Direct messages often require a thoughtful reply drawing on context from other sources, like a CRM record, a project doc, or a previous email thread. Your AI assistant can pull that context and draft a reply to the DM via im:write. You review the draft before sending, or configure it to post directly for routine replies.

Best practices for Slack MCP permissions

A common question from solo operators: how much access should I grant? The answer is always the minimum needed for your specific workflow.

Recommended Slack permissions to grant

For a basic read, search, and post workflow, grant only these scopes:

• channels:read - lets the AI list your public channels

• channels:history - lets the AI read message content in channels it has access to

• chat:write - lets the AI post messages to channels

• users:read - lets the AI look up workspace members by name or ID

For DM workflows, add im:write and im:history. Grant only what your specific workflow requires.

Preventing over-privileged access

Configure Slack's built-in channel posting permissions to restrict which channels any bot can post into. This is managed at the workspace level inside Slack's channel settings, and it gives you a second layer of control on top of what you grant during OAuth.

Start with a free Composio account, connect Slack in your next 10 minutes, and run your first AI Slack query before you close this tab. Composio's MCP documentation covers every other toolkit you can add from there, including Gmail, Notion, and GitHub setups, each following the same single-session connection pattern.

FAQs

Which AI assistants support Slack MCP?

Claude Desktop, ChatGPT, and Cursor support Slack MCP directly through Composio's managed gateway. You connect them by adding a single server URL to your assistant's configuration file, with no additional packages required.

How do you manage Slack visibility for AI posts?

Channel-level posting restrictions are managed through Slack's built-in channel permissions settings in your workspace. You control which channels any connected app can post into directly from within Slack, giving you fine-grained control independent of the OAuth scopes you grant.

Is Slack MCP secure?

Yes. Composio is SOC 2 Type II certified and stores credentials in a managed token store rather than local files. For users who want additional privacy, the Zero Data Retention add-on prevents request and response payloads from being stored at all.

What Slack actions can the AI assistant perform?

The Composio Slack toolkit supports posting messages, reading channel history, searching the workspace, listing users and channels, managing reactions, sending direct messages, and creating reminders, among others. You enable only the specific actions your workflow requires.

Key terms glossary

Model Context Protocol (MCP): A universal connector that allows AI assistants to securely read data from and write data to other software tools like Slack, Gmail, or Notion.

OAuth token: A secure digital key that allows your AI assistant to perform actions in Slack on your behalf without exposing your password.

Zero Data Retention (ZDR): A security setting that prevents integration platforms from storing the content of your API requests and responses, available as an add-on on Composio's Pro tier and above.

Scope: A specific permission granted to an OAuth app that defines exactly what it can and can't do inside your Slack workspace. For example, chat:write allows posting but not reading history.

Connect Link: A single URL generated by Composio that lets a user authenticate a new app connection in one click, giving your AI assistant immediate access to that service's action toolkit without any code changes on your end.