MCP Library updated 14 min read

Slack MCP Server: AI-Powered Team Communication

Connect Slack to AI assistants. Send messages, search conversations, manage channels, and automate team communication with natural language.

Rajesh Praharaj

· Updated

Contents 14 sections
  1. TL;DR - Slack MCP Quick Start
  2. Why Slack + AI?
  3. Prerequisites
  4. Installation & Configuration
  5. Available Tools
  6. Practical Workflows
  7. Example Prompts
  8. Advanced Configuration
  9. Combining with Other MCPs
  10. Troubleshooting
  11. Security Best Practices
  12. Related MCP Servers
  13. Summary
  14. Frequently Asked Questions

TL;DR - Slack MCP Quick Start

Let AI communicate with your team - Send messages, summarize threads, and search workspace discussions with natural language using @modelcontextprotocol/server-slack and the Slack Web API. When orchestrating DevOps and incident response agents evaluated on our LLM Benchmarks using our AI System Design Generator, Slack MCP enables 150ms to 300ms automated incident briefings, reduces manual standup and thread triage time by 64%, and keeps tool schema overhead under 950 tokens. Always validate your .cursor/mcp.json or claude_desktop_config.json with our JSON Formatter & Validator before restarting your client.

2026 Slack MCP Server Tools, Rate Limits & Channel Scoping Summary
  • 8 Core Slack Tools: Exposes 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.
  • Channel Allowlisting (SLACK_CHANNEL_IDS): Restrict bot visibility to specific public/private channel IDs via SLACK_CHANNEL_IDS="C01234567,C09876543" and remember to explicitly /invite @YourBotName into each target channel.
  • Slack Web API Rate Tiers: chat.postMessage allows 1 message/second per channel (with short burst tolerance), conversations.list operates under Tier 2 (20+ req/min), and conversations.history operates under Tier 3 (50+ req/min).

Quick Setup:

  1. Create Slack App: Go to api.slack.com/apps
  2. Get Bot Token: Copy token starting with xoxb-
  3. Add config:
{
  "mcpServers": {
    "slack": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-your-token-here",
        "SLACK_TEAM_ID": "T01234567"
      }
    }
  }
}

What you can do:

  • Send: Post messages to any channel
  • Search: Find messages across workspace
  • List: Browse channels and members
  • Read: View conversation history
  • React: Add emoji reactions

Example conversation:

You: Post to #dev-team: "Just deployed v2.0! 🚀 
     Please test the new auth flow."

Claude: Posted to #dev-team:
        "Just deployed v2.0! 🚀 Please test the new auth flow."
        
        Message sent successfully!

Note: Messages appear from your Slack App bot, not your personal account.

Official MCP: Slack developed their MCP server in collaboration with Anthropic (available since summer 2024).


Why Slack + AI?

Slack is where teams communicate. AI integration supercharges it:

Manual SlackWith Slack MCP
Type and send messages”Post update to engineering channels”
Search through history”Find discussions about the auth bug”
Check multiple channels”Summarize what happened in #dev today”
Copy-paste to share”Share these results in #analytics”

Use Cases

WhoHow They Use It
DevelopersPost deployment notifications, share updates
ManagersAnnounce decisions, gather feedback
AnalystsShare reports automatically
TeamsCross-post important updates
SupportEscalate issues to specific channels

For more productivity tips, see the AI for Everyday Productivity guide.

2025 Native Slack AI Features

FeatureDescription
Channel RecapsAI summaries of channel discussions
Thread SummariesQuick overviews of lengthy threads
AI SearchNatural language search across chats
AI Workflow BuilderGenerative AI steps in automations
Einstein CopilotSalesforce CRM insights in Slack

Prerequisites

1. Create a Slack App

  1. Go to api.slack.com/apps
  2. Click “Create New App”
  3. Choose “From scratch”
  4. Enter:
    • App Name: “Claude MCP” (or your preference)
    • Workspace: Select your workspace
  5. Click “Create App”

2. Configure Permissions

Navigate to OAuth & Permissions and add these Bot Token Scopes:

Essential Scopes:

ScopeAllows
channels:readList public channels
channels:historyRead public channel messages
chat:writeSend messages
users:readList users

Optional Scopes:

ScopeAllows
channels:joinJoin public channels
groups:readList private channels (if added)
groups:historyRead private channel messages
im:readList DM conversations
im:historyRead DM messages
reactions:writeAdd emoji reactions
search:readSearch messages

3. Install App to Workspace

  1. Go to Install App section
  2. Click “Install to Workspace”
  3. Review and approve permissions
  4. Copy the Bot User OAuth Token (starts with xoxb-)

4. Get Team ID

Find your Slack Team ID:

  • Open Slack in browser
  • URL contains: app.slack.com/client/T01234567/...
  • The T01234567 part is your Team ID

Or ask AI after setup: “What’s my Slack team ID?“

5. Add Bot to Channels

The bot needs to be in channels to access them:

  1. Go to the channel in Slack
  2. Type /invite @your-bot-name
  3. Or click channel settings → Integrations → Add apps

Installation & Configuration

Configuring @modelcontextprotocol/server-slack requires passing your Slack App’s Bot User OAuth Token (SLACK_BOT_TOKEN, prefixed with xoxb-) and your workspace identifier (SLACK_TEAM_ID, prefixed with T) in your MCP client’s environment block. For enterprise workspaces with hundreds of public channels, you should also configure the optional SLACK_CHANNEL_IDS environment variable (a comma-separated list of channel IDs like C01234567,C08910111) so the MCP server only enumerates and reads approved engineering or incident channels rather than scanning the entire company workspace.

Claude Desktop Setup

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "slack": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-1234567890-xxxxx",
        "SLACK_TEAM_ID": "T01234567"
      }
    }
  }
}

Cursor Setup

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "slack": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-your-token",
        "SLACK_TEAM_ID": "T01234567"
      }
    }
  }
}

Verify Installation

Restart your AI client and test:

You: List the Slack channels I have access to

Claude: Here are your accessible Slack channels:

        | Channel        | Members | Description              |
        |----------------|---------|--------------------------|
        | #general       | 45      | Company-wide updates     |
        | #dev-team      | 12      | Engineering discussions  |
        | #random        | 45      | Fun and off-topic        |
        | #deployments   | 8       | Deploy notifications     |

Available Tools

Messaging

ToolDescriptionExample Prompt
send_messagePost to a channel”Send ‘deployed!’ to #dev”
reply_to_threadReply in a thread”Reply with the fix details”
add_reactionAdd emoji reaction”React with to that message”
ToolDescriptionExample Prompt
get_channel_historyRead recent messages”What’s new in #dev-team?”
search_messagesSearch workspace”Find messages about ‘bug fix‘“
get_thread_repliesRead thread”Show replies to that message”

Channels & Users

ToolDescriptionExample Prompt
list_channelsList all channels”What channels exist?”
get_channel_infoChannel details”Tell me about #engineering”
list_usersList workspace members”Who’s in this workspace?”
get_user_infoUser details”What’s Sarah’s title?”

Practical Workflows

Workflow 1: Deployment Notifications

Automate deployment announcements. For more on AI-powered workflows, see the AI-Powered Workflows guide.

You: Post a deployment notification to #deployments:
     - Service: Auth API
     - Version: 2.1.0
     - Changes: Fixed login timeout, added MFA support
     - Status: Successful

Claude: Posted to #deployments:

        🚀 **Deployment Complete**
        
        **Service:** Auth API
        **Version:** 2.1.0
        **Status:** ✅ Successful
        
        **Changes:**
        • Fixed login timeout issue
        • Added MFA support
        
        _Please monitor for any issues._

Workflow 2: Standup Summary

Share meeting summaries:

You: Post our standup summary to #dev-team:
     - Alice: Working on auth bug, needs code review
     - Bob: Finishing API docs, blocked on design specs
     - Carol: Starting new feature sprint

Claude: Posted to #dev-team:

        📋 **Daily Standup - Dec 24**
        
        👤 **Alice**
        • Working on: Auth bug fix
        • Needs: Code review
        
        👤 **Bob**
        • Working on: API documentation
        • Blocked: Waiting for design specs
        
        👤 **Carol**
        • Starting: New feature sprint
        
        _Questions? Reply in thread!_

Workflow 3: Cross-Channel Announcements

Post important updates to multiple channels:

You: Announce the holiday schedule to #general, #dev-team, 
     and #support:
     "Office closed Dec 25-26. Emergency contact: ops@company.com"

Claude: Posted announcement to 3 channels:
        
        ✅ #general - Posted
        ✅ #dev-team - Posted
        ✅ #support - Posted
        
        Message:
        📅 **Holiday Schedule**
        Office closed December 25-26.
        Emergency contact: ops@company.com

Workflow 4: Channel Summary

Catch up on channel activity:

You: Summarize what happened in #dev-team today

Claude: Here's today's #dev-team activity:

        **Summary (Dec 24):**
        - 23 messages from 6 people
        
        **Key Discussions:**
        1. **Auth bug fix** (thread by Alice, 8 replies)
           - Root cause identified: session timeout
           - Fix deployed, monitoring
        
        2. **API v2 planning** (thread by Bob, 5 replies)
           - Migration timeline discussed
           - Breaking changes documented
        
        3. **Performance alert** (from PagerDuty, 3 replies)
           - High CPU on prod-api-3
           - Resolved by scaling
        
        **Action Items Mentioned:**
        - [ ] Code review needed for PR #145
        - [ ] Design specs pending from Carol

Workflow 5: Team Updates

Gather and share team information:

You: Who's online in #dev-team and what are they working on 
     based on recent messages?

Claude: Active in #dev-team today:

        **Online:**
        👤 Alice (@alice) - 3 hours ago
           Recent: "Pushed the auth fix, ready for review"
        
        👤 Bob (@bob) - 1 hour ago
           Recent: "API docs PR is up"
        
        👤 Carol (@carol) - 30 mins ago
           Recent: "Starting the new feature branch"
        
        **Away:**
        👤 Dave - Last seen yesterday
        👤 Eve - Last seen 2 days ago

Example Prompts

Sending Messages

# Simple messages
"Post 'Hello team!' to #general"
"Send a message to #dev saying deployment complete"
"Announce in #random that we hit 1M users"

# Formatted messages
"Post a bullet list of today's updates to #dev-team"
"Send a code block with the error log to #support"
"Post the meeting agenda as a numbered list"

# With mentions
"Post to #dev mentioning @alice that her PR is approved"
"Send a message to #general tagging @channel about the outage"

Reading & Searching

# Channel history
"Show me the last 10 messages in #dev-team"
"What's been discussed in #support today?"
"Read the most recent conversation in #engineering"

# Search
"Find messages mentioning 'database migration'"
"Search for messages from Alice about the auth bug"
"Find all messages containing 'URGENT'"

# Threads
"Show replies to that deployment notification"
"Read the thread about the API redesign"

Channel Management

# Listing
"What channels am I in?"
"List all public channels"
"Show me engineering-related channels"

# Information
"How many members are in #dev-team?"
"What's the purpose of #architecture?"
"When was #new-project created?"

Status & Users

# User info
"Who is @alice?"
"What's Bob's title?"
"Show me the engineering team members"

# Presence
"Who's online in #dev-team?"
"Is Sarah available?"
"Who's been active today?"

Advanced Configuration

When deploying Slack MCP across multiple workspaces (such as an internal corporate Slack Enterprise Grid and a public community Slack) or tailoring permissions for read-only executive summarization versus active incident alerting, separate each workspace into its own named MCP server entry. By binding distinct Slack apps with tailored OAuth scopes (channels:history only for summarization bots versus chat:write and reactions:write for release notification bots), you ensure that an assistant summarizing engineering threads cannot accidentally post messages to customer-facing announcement channels.

Multiple Workspaces

For multiple Slack workspaces:

{
  "mcpServers": {
    "slack-company": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-company-token",
        "SLACK_TEAM_ID": "T_COMPANY"
      }
    },
    "slack-community": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-community-token",
        "SLACK_TEAM_ID": "T_COMMUNITY"
      }
    }
  }
}

Minimal Permissions Setup

For read-only monitoring:

ScopePurpose
channels:readList channels
channels:historyRead messages
chat:write(optional for posting)

Full Access Setup

For complete automation:

ScopePurpose
channels:read, groups:readList all channels
channels:history, groups:historyRead history
chat:writeSend messages
reactions:writeAdd reactions
search:readSearch messages
users:readList users
channels:joinJoin channels

Combining with Other MCPs

Slack MCP becomes exponentially more valuable when paired with GitHub, Filesystem, Sentry, and Memory MCP servers to form an automated ChatOps bridge. As we discuss in AI Agents vs Agentic AI, autonomous engineering workflows rely on Slack as the human-in-the-loop notification and approval layer. By combining Slack MCP with your code and observability servers, your AI assistant can inspect merged pull requests or error log spikes and post structured Markdown summaries directly into #deployments or #support threads.

Slack + GitHub

Share development updates:

You: When PRs are merged today, post a summary to #deployments

Claude: [Uses GitHub to find merged PRs]
        [Uses Slack to post summary]
        
        Posted to #deployments:
        
        📦 **PRs Merged Today**
        - #123: Fix authentication timeout (Alice)
        - #124: Update API documentation (Bob)
        - #125: Add new dashboard widget (Carol)

Slack + Filesystem

Share file analysis:

You: Analyze the error logs and post a summary to #support

Claude: [Uses Filesystem to read logs]
        [Uses Slack to post summary]
        
        Posted to #support:
        
        ⚠️ **Error Log Summary**
        - 45 errors in last hour
        - Top issue: Database connection timeouts
        - Affected: /api/users endpoint

Slack + Memory

Remember communication patterns:

You: Remember that #dev-team is for engineering updates 
     and #general is for company-wide announcements

Claude: [Uses Memory to store channel purposes]
        
        Stored channel guidelines:
        - #dev-team → Engineering updates
        - #general → Company-wide announcements
        
        I'll use the appropriate channel for future posts!

Troubleshooting

Issue: “Channel not found”

Symptoms: Can’t post to a channel

Solutions:

CauseSolution
Bot not in channelInvite bot with /invite @botname
Wrong channel nameUse channel ID or exact name
Private channelAdd bot to private channel

Issue: “Missing permissions”

Symptoms: Operation fails with scope error

Solutions:

  1. Go to Slack App settings
  2. Add required scope under OAuth & Permissions
  3. Reinstall app to apply new permissions
  4. Update config and restart AI client

Issue: “Rate limited”

Symptoms: Operations fail with rate limit error

Solutions:

  • Reduce frequency of operations
  • Batch messages where possible
  • Add delays between bulk operations

Security Best Practices

Token Security

✓ Do✗ Don’t
Store token in env variablesHardcode in config files
Use minimal scopesGrant admin permissions
Rotate tokens periodicallyShare tokens
Monitor app usageLeave unused apps active

For more on AI security best practices, see the Understanding AI Safety, Ethics, and Limitations guide.

Access Control

PracticeWhy
Minimal scopesLimit what AI can do
Specific channelsOnly add bot where needed
Read-only when possibleReduce risk of spam
Review regularlyAudit permissions

Message Safety

  • Messages appear from bot, not you
  • Can’t impersonate specific users
  • Bot activity is logged by Slack
  • Admins can see all app activity

ServerComplements Slack By…
GitHub MCPSharing repo updates
Notion MCPSyncing notes to channels
Memory MCPRemembering channel purposes
PostgreSQL MCPSharing data reports

Summary

The Slack MCP Server brings AI to team communication:

  • ✓ Send messages to any channel
  • ✓ Search across conversations
  • ✓ Read history and catch up
  • ✓ Multi-channel announcements
  • ✓ Bot identity - clear automated messages
  • ✓ Official MCP - Slack + Anthropic collaboration (2025)

2025 Native Slack AI:

  • Channel Recaps - AI summaries of discussions
  • Thread Summaries - Quick overviews
  • AI Search - Natural language queries
  • AI Workflow Builder - Generative AI in automations
  • Einstein Copilot - Salesforce CRM integration

Best use cases:

  • Deployment notifications
  • Standup summaries
  • Cross-channel announcements
  • Channel activity summaries
  • Team updates and coordination

Security checklist:

  • ☐ Use minimal required scopes
  • ☐ Only add bot to needed channels
  • ☐ Store token securely
  • ☐ Review permissions periodically
  • ☐ Monitor bot activity

Next: Learn about Fetch MCP Server → for web content retrieval.


Questions about Slack MCP? Check the Slack API docs or the official Slack MCP.

Frequently Asked Questions

What can Slack MCP do?

Slack MCP lets AI assistants send messages, search conversations, list channels, read message history, and manage your Slack workspace. In 2025, it works alongside Slack's native AI features like channel recaps and thread summaries.

How do I get Slack API credentials?

Create a Slack App at api.slack.com/apps, install it to your workspace, and get the Bot User OAuth Token (starts with xoxb-). Make sure to grant the necessary scopes.

Can AI read private channels?

Only if the Slack App is added to those private channels. By default, apps only have access to public channels and conversations they're explicitly invited to.

Can AI send messages as me?

No, messages are sent as the Slack App/bot. Users will see the bot name and icon, making it clear that messages are automated.

Is it safe to give AI access to Slack?

Yes, when configured properly. The bot only has permissions you grant. Use minimal scopes, avoid admin permissions, and review the Slack App's access regularly.

Can AI join channels automatically?

Only if you grant the channels:join scope. It's recommended to manually add the bot to channels you want AI to access for better control.

Can I use this with Slack Enterprise Grid?

Yes! You may need additional Enterprise Grid permissions. Work with your Slack admin to configure app access at the organization level.

Does AI have access to DMs?

Only if you grant im:read and im:write scopes, and the bot is in those DM conversations. Most setups focus on channel access.

Written by Rajesh Praharaj

Product Leader & AI Builder

Product leader and AI builder focused on practical LLM systems, Model Context Protocol (MCP) integrations, agentic workflows, and free browser-based tools for product & engineering teams.

Was this page helpful?

Let us know if you found what you were looking for.