Build a Codebase-Aware Slack Bot with Visor

This guide walks you through creating a Slack bot that can answer questions about your codebase. Your team can ask questions in Slack and get intelligent answers grounded in your actual code.

Time to complete: ~10 minutes

What You'll Build

A Slack bot that:

• Responds to mentions and direct messages
• Uses Probe to search and understand your codebase
• Provides accurate, context-aware answers about your code
• Works in channels, DMs, and group messages

Prerequisites

• A Slack workspace where you have permission to create apps
• An Anthropic API key (or other LLM provider)
• Visor installed (npm install -g @probelabs/visor or via npx)
• Your codebase accessible locally or via git

Step 1: Create a Slack App

1.1 Go to Slack App Creation

1. Visit api.slack.com/apps
2. Click Create New App
3. Select From an app manifest
4. Choose your workspace

1.2 Use the App Manifest

Copy and paste this manifest in the JSON tab:

{
    "display_information": {
        "name": "Probe",
        "description": "A friendly bot to answer questions about code",
        "background_color": "#000000"
    },
    "features": {
        "app_home": {
            "home_tab_enabled": true,
            "messages_tab_enabled": false,
            "messages_tab_read_only_enabled": false
        },
        "bot_user": {
            "display_name": "Probe",
            "always_online": true
        }
    },
    "oauth_config": {
        "redirect_urls": [
            "https://localhost:3993/oauth/callback"
        ],
        "scopes": {
            "user": [
                "search:read.public"
            ],
            "bot": [
                "app_mentions:read",
                "channels:history",
                "chat:write",
                "groups:history",
                "groups:read",
                "im:history",
                "im:read",
                "im:write",
                "mpim:history",
                "reactions:read",
                "reactions:write",
                "users:read.email",
                "users:read"
            ]
        }
    },
    "settings": {
        "event_subscriptions": {
            "bot_events": [
                "app_mention",
                "message.channels",
                "message.groups",
                "message.im",
                "message.mpim"
            ]
        },
        "interactivity": {
            "is_enabled": true
        },
        "org_deploy_enabled": false,
        "socket_mode_enabled": true,
        "token_rotation_enabled": false
    }
}

You can customize the display_name to whatever you'd like your bot to be called in Slack.

Click Create to create the app.

Understanding the Manifest

Section	Purpose
display_information	Bot name and appearance in Slack
bot_user	The bot's display name and online status
oauth_config.scopes	Permissions the bot needs to read messages and respond
event_subscriptions.bot_events	Events that trigger the bot (mentions, messages)
socket_mode_enabled	Enables real-time communication without a public URL

Step 2: Get Your Tokens

After creating the app, you need two tokens:

2.1 Bot Token (SLACK_BOT_TOKEN)

1. Go to OAuth & Permissions in the left sidebar
2. Click Install to Workspace
3. Authorize the app
4. Copy the Bot User OAuth Token (starts with xoxb-)

2.2 App Token (SLACK_APP_TOKEN)

1. Go to Basic Information in the left sidebar
2. Scroll to App-Level Tokens
3. Click Generate Token and Scopes
4. Name it socket-mode or similar
5. Add the scope connections:write
6. Click Generate
7. Copy the token (starts with xapp-)

Step 3: Create the Visor Workflow

Create a file called slack-bot.yaml in your project:

steps:
  ask:
    type: human-input
    prompt: |
      Hi! Ask me something.

  checkout:
    type: git-checkout
    depends_on: [ask]
    ref: "main"
    repository: "probelabs/probe"

  answer:
    type: ai
    depends_on: [ask, checkout]
    prompt: |
      User asked: {{ outputs['ask'].text }}
      Repo path: {{ outputs['checkout'].path }}
      Reply in 1–3 sentences.

Step 4: Set Environment Variables

Create a .env file (add to .gitignore!):

# Slack tokens (from Step 2)
SLACK_BOT_TOKEN=xoxb-your-bot-token-here
SLACK_APP_TOKEN=xapp-your-app-token-here

# LLM API key
ANTHROPIC_API_KEY=sk-ant-your-key-here

# Optional: Use a different model provider
# OPENAI_API_KEY=sk-your-key-here

Or export them in your shell:

export SLACK_BOT_TOKEN="xoxb-..."
export SLACK_APP_TOKEN="xapp-..."
export ANTHROPIC_API_KEY="sk-ant-..."

Step 5: Run the Bot

Run with npx (no install needed):

npx -y @probelabs/visor@latest --config slack-bot.yaml --slack

You should see:

✓ Connected to Slack
✓ Listening for events...

Test It

1. Go to your Slack workspace
2. Invite the bot to a channel: /invite @Probe
3. Mention the bot with a question: @Probe how does authentication work?

The bot will:

1. Search your codebase using Probe
2. Find relevant code
3. Formulate an answer grounded in what it found
4. Reply in the thread

Step 6: Deploy to Production

Self-Host with Docker

FROM node:20-slim

WORKDIR /app
COPY slack-bot.yaml .

RUN npm install -g @probelabs/visor

CMD ["npx", "-y", "@probelabs/visor@latest", "--config", "slack-bot.yaml", "--slack"]

docker build -t slack-bot .
docker run -d --name slack-bot \
  -e SLACK_BOT_TOKEN=xoxb-... \
  -e SLACK_APP_TOKEN=xapp-... \
  -e ANTHROPIC_API_KEY=sk-ant-... \
  slack-bot

Run on Your Infrastructure

Use systemd, PM2, or your preferred process manager:

# With PM2
pm2 start "npx -y @probelabs/visor@latest --config slack-bot.yaml --slack" --name slack-bot

Troubleshooting

Bot Not Responding

1. Check the bot is running: Look for "Connected to Slack" in logs
2. Verify tokens: Ensure both SLACK_BOT_TOKEN and SLACK_APP_TOKEN are set
3. Check permissions: The bot needs to be invited to channels
4. Socket mode: Ensure socket mode is enabled in Slack app settings

"Not Authorized" Errors

• Reinstall the app to your workspace (OAuth & Permissions > Install to Workspace)
• Verify the bot token starts with xoxb-
• Verify the app token starts with xapp-

Bot Responds Slowly

• Probe searches can take a few seconds on large codebases
• Consider narrowing the search paths in context.paths
• Use context.exclude to skip irrelevant directories

Empty or Unhelpful Answers

• Check that your codebase paths are correct
• Verify Probe can search your code: probe search "your query" ./src
• Add more specific instructions in the system prompt

Environment Variables Reference

Variable	Required	Description
SLACK_BOT_TOKEN	Yes	Bot User OAuth Token (xoxb-...)
SLACK_APP_TOKEN	Yes	App-Level Token with connections:write (xapp-...)
ANTHROPIC_API_KEY	Yes*	Anthropic API key for Claude
OPENAI_API_KEY	Alt*	OpenAI API key (alternative to Anthropic)

*One LLM provider API key is required.

---

Questions? Join our Discord or book a demo.