Slack is where most teams talk through decisions, escalate problems, and share project context. If you want your AI assistant to read those conversations directly, a Slack MCP server connects them. You configure it once, and Claude can summarize threads, search for decisions made about a project, or retrieve what the team agreed on last Tuesday without you copying and pasting anything. MCPFind indexes 96 servers in the communication category. This guide covers creating the Slack bot, setting the right scopes, connecting it in Claude Desktop, and what you can ask once it is live. For MCP background, start with What Is MCP?.
What Does a Slack MCP Server Do?
A Slack MCP server connects your AI client to the Slack API so it can read channel messages, search workspace history, list members, and retrieve thread replies. Instead of copying Slack conversations into Claude's chat window, you ask Claude to look them up directly.
The server runs locally on your machine using a Slack bot token for authentication. It only has access to channels the bot has been invited to. The tools it exposes typically include reading recent channel messages, searching across the workspace, getting thread context, and listing channels the bot can see.
What most Slack MCP setups do not do by default: post messages. Read-only access is the standard starting configuration, and it is also the safer one. If you want the AI to be able to write to Slack, you add the write scope intentionally after you have seen how it uses read access. Starting narrow and expanding deliberately is the right pattern for any MCP server that touches team communication data.
What You Need Before Setting Up Slack MCP
You need a Slack workspace, a Slack app, and the right permission scopes. Creating the app takes about five minutes at api.slack.com.
Go to api.slack.com/apps and create a new app from scratch. Name it something recognizable; your team will see this as the bot's display name in channels. Navigate to OAuth & Permissions and add these bot token scopes: channels:history, channels:read, groups:history, groups:read, search:read, and users:read. These give read access to public channels and any private channels the bot is later invited to.
Install the app to your workspace. Slack generates a Bot User OAuth Token that starts with xoxb-. Copy that token and keep it safe; you will paste it into the MCP config in the next step. If your workspace requires admin approval for app installations, a workspace admin needs to approve it before the token is generated.
How to Add Slack MCP to Claude Desktop
Open your Claude Desktop config at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS. Windows users will find it at %APPDATA%\Claude\claude_desktop_config.json. Add the Slack server under mcpServers:
{
"mcpServers": {
"slack": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-slack"],
"env": {
"SLACK_BOT_TOKEN": "xoxb-your-token-here",
"SLACK_TEAM_ID": "T0123456789"
}
}
}
}Find your Team ID in the Slack workspace URL or under Settings in your workspace admin panel. Save the config, then quit and relaunch Claude Desktop. A hammer icon in Claude's interface confirms the Slack tools loaded. If the connection fails, check that the bot was installed to the workspace and that the token scopes match what was listed in the previous step.
The same config block works in Cursor, Windsurf, and VS Code. Add it under mcpServers in your editor's MCP config file and the server runs without modification.
What Can You Ask Claude With Slack Connected?
Once the server is live, you can ask Claude things that would normally require scrolling through multiple channels:
- "Summarize the last 30 messages from #engineering" condenses a busy thread without you reading it first.
- "What did we decide about the Q2 roadmap?" searches your workspace and returns relevant messages with context.
- "Did anyone mention the API rate limit issue this week?" pulls matching messages with timestamps and author names.
The server only reads what the bot can see. Start by inviting the bot to your two or three most active project channels, then expand from there after you observe how the agent uses the access. That gives you control over context scope rather than granting broad workspace visibility from the start. For teams that also use automation tools, the automation category on MCPFind has 31 servers that pair well with Slack MCP for trigger-based workflows.
For a similar connection with a document-based tool, the guide for connecting Notion to Claude with MCP covers the same pattern for knowledge management workflows.
Adam Bush