Every team has questions that eat up time. "Where's that document?" "What did we decide in the last meeting?" "How do I use this internal tool?" An AI bot in Slack can answer these instantly, freeing your team to focus on real work.

Unlike Slack's built-in AI features, a custom bot powered by Claude gives you full control: custom knowledge, specific behaviors, and integration with your existing systems.

This guide shows you how to set up a Claude-powered AI assistant in your Slack workspace.

Why Add AI to Slack?

AI assistants transform how teams find and share information

What a Slack AI bot can do:

• Answer questions using company knowledge
• Summarize long threads and channels
• Draft responses to common questions
• Search across channels and documents
• Translate between languages
• Onboard new team members faster

Real impact:

• Reduce repetitive questions to specific people
• Faster information discovery
• Consistent answers to common questions
• Available 24/7, no waiting for colleagues

The best Slack bots feel like a helpful team member who knows everything and never gets tired.

Architecture Overview

How the AI connects to your Slack workspace

Components:

1. Slack App - The bot presence in your workspace
2. OpenClaw or custom backend - Processes messages with Claude
3. Claude API - Provides AI responses
4. Knowledge base (optional) - Custom documents and context

Message flow:

1. User mentions bot or DMs it
2. Slack sends message to your backend
3. Backend sends to Claude with context
4. Claude response sent back to Slack

Step 1: Create Your Slack App

Setting up your Slack app

Create the app:

1. Go to api.slack.com/apps
2. Click "Create New App"
3. Choose "From scratch"
4. Name it (e.g., "Claude Assistant")
5. Select your workspace

Configure bot permissions:

Navigate to OAuth & Permissions → Scopes → Bot Token Scopes:

Required scopes:
- app_mentions:read (respond to mentions)
- chat:write (send messages)
- channels:history (read public channels)
- groups:history (read private channels you're invited to)
- im:history (read DMs)
- users:read (user info for context)

Enable Events:

Navigate to Event Subscriptions:

• Toggle "Enable Events" on
• Add Request URL (your backend endpoint)
• Subscribe to bot events: app_mention, message.im

Install to workspace:

• OAuth & Permissions → Install to Workspace
• Authorize the permissions
• Copy the Bot User OAuth Token (starts with xoxb-)

Step 2: Connect to OpenClaw

Connecting OpenClaw to your Slack workspace

OpenClaw handles Slack integration natively:

Add Slack channel:

openclaw config set slack.botToken "xoxb-your-token"
openclaw config set slack.signingSecret "your-signing-secret"
openclaw config set slack.enabled true

openclaw gateway restart

Configure behavior:

// openclaw.json
{
  "channels": {
    "slack": {
      "enabled": true,
      "botToken": "xoxb-...",
      "signingSecret": "...",
      "responseMode": "thread",  // Reply in threads
      "allowedChannels": ["*"],  // Or specific channel IDs
      "systemPrompt": "You are a helpful team assistant."
    }
  }
}

Test the connection:

openclaw channels status

You should see Slack listed as connected. Now mention your bot in Slack and watch it respond.

Step 3: Add Custom Knowledge

Adding company-specific knowledge

A generic AI is helpful, but a bot with company knowledge is transformative.

Option 1: System prompt context

Add key information to your system prompt covering: company products with common questions, internal processes (PTO, expenses) with links, and team routing (which channels/people handle what topics). Include guidance to refer users to official channels for approvals.

Option 2: File-based knowledge

Add documents to OpenClaw's workspace:

~/.openclaw/
├── knowledge/
│   ├── onboarding.md
│   ├── product-faq.md
│   ├── engineering-standards.md
│   └── team-directory.md

Reference in system prompt:

You have access to company documentation in the knowledge/ folder.
Search these files to answer questions about company policies and processes.

Option 3: Retrieval augmented generation (RAG)

For larger knowledge bases, implement proper RAG:

• Index documents with embeddings
• Search relevant content for each query
• Include in context for responses

Step 4: Configure Team Workflows

Automating team collaboration

Thread summaries:

@claude-bot summarize this thread

Configure the bot to recognize summary requests and provide concise overviews.

Scheduled updates:

openclaw cron add "standup-reminder" "0 9 * * 1-5" \
  "Post to #engineering: Time for daily standup! 
   What did you do yesterday? What are you doing today? Any blockers?"

Channel monitoring:

openclaw cron add "support-check" "0 */2 * * *" \
  "Check #support for questions without responses in the last 2 hours.
   If found, suggest answers or escalate to relevant team members."

Auto-responses:

Configure patterns for common questions:

• "Where is X?" → Link to documentation
• "How do I Y?" → Step-by-step instructions
• "Who handles Z?" → Point to right person/channel

Step 5: Security and Access Control

Protect sensitive information appropriately

Channel restrictions:

{
  "channels": {
    "slack": {
      "allowedChannels": ["C123ABC", "C456DEF"],
      "blockedChannels": ["C789GHI"]
    }
  }
}

User permissions:

{
  "channels": {
    "slack": {
      "allowedUsers": ["U123", "U456"],
      "adminUsers": ["U789"]
    }
  }
}

Sensitive topics:

Add to system prompt:

Never discuss or reveal:
- Salary information
- Personal performance reviews
- Confidential business strategy
- Individual user data

For these topics, direct users to HR or appropriate channels.

Best Practices

Keys to successful Slack bot deployment

Naming and personality:

• Give it a memorable name
• Define consistent tone (professional, friendly, etc.)
• Make it clear it's AI, not human

Response quality:

• Keep responses concise for chat context
• Use formatting (lists, bold) appropriately
• Offer to elaborate when relevant

Escalation:

• Know when to hand off to humans
• Point to the right channels/people
• Don't pretend to know what it doesn't

Continuous improvement:

• Review bot interactions weekly
• Update knowledge base regularly
• Add common questions to FAQ

Conclusion

Empowering your team with AI

A well-configured Slack AI bot becomes an invaluable team resource. It handles repetitive questions, provides instant access to information, and frees your team to focus on work that matters.

Start simple—basic Q&A with some company context. Then expand based on what your team actually uses. The best bot configurations evolve from real usage patterns.

Get started:

• OpenClaw setup for the backend
• Discord bot setup for similar process
• Automation guide for scheduled tasks

Your team deserves better than repeating the same answers. Give them an AI assistant.

FAQ

Common questions about Slack AI bots

Is this different from Slack's built-in AI?

Yes. Slack's AI uses their models with their features. A custom bot using Claude gives you control over behavior, knowledge, and integrations.

Can the bot access private channels?

Only if explicitly invited. It respects Slack's permission model.

How do I prevent sensitive data leakage?

Restrict channel access, configure topic blockers in system prompt, and review conversation logs periodically.

What about API costs?

Depends on usage. A team of 20 with moderate usage might cost $20-50/month in Claude API fees. Heavy usage scales up proportionally.

Can multiple workspaces share one bot?

Yes, with enterprise setup. Each workspace needs its own app installation, but they can share the same backend.