Get LocalClaw running in 5 minutes!

# Check Node.js (need 18+)
node --version

# Check Ollama is running
curl http://localhost:11434/api/tags

# If Ollama isn't running:
ollama serve

# From the localclaw directory:
npm install
npm run build
npm link

# Run the setup wizard
localclaw onboard

# Pull a lightweight model (if you don't have one)
ollama pull qwen3:4b

# Verify everything works
localclaw doctor

# Create a simple file
localclaw agent "Create a file called hello.txt with the text 'Hello from LocalClaw!'"

# Check the result
cat ~/localclaw/workspace/hello.txt

# Generate a Python script
localclaw agent "Create a Python script called fibonacci.py that calculates the first 10 Fibonacci numbers and prints them"

# Run it!
python ~/localclaw/workspace/fibonacci.py

# List all jobs
localclaw jobs list

# Show details of the most recent job
localclaw jobs show <job-id-from-list>

# Make sure you ran npm link
cd /path/to/localclaw
npm link

# Or use npx
npx tsx src/cli/index.ts onboard

# Start Ollama in a separate terminal
ollama serve

# Or check if it's running
ps aux | grep ollama

# Pull the default model
ollama pull qwen3:4b

# Or list what you have
ollama list

All operations are restricted to ~/localclaw/workspace by default for safety. Check that your task is creating/reading files in the workspace.

1. Read the examples: Check out EXAMPLES.md for more complex use cases
2. Customize config: Edit ~/.smallclaw/config.json to adjust:
   • Which model to use
   • Tool permissions
   • Workspace location
3. Try different models: Experiment with qwen2.5-coder:32b or llama-3.3:70b
4. Build skills: Create custom SKILL.md files for repeated tasks

{
  "models": {
    "primary": "qwen3:4b"
  },
  "ollama": {
    "concurrency": {
      "llm_workers": 1,
      "tool_workers": 2
    }
  }
}

{
  "models": {
    "primary": "qwen2.5-coder:32b"
  },
  "ollama": {
    "concurrency": {
      "llm_workers": 1,
      "tool_workers": 3
    }
  }
}

{
  "models": {
    "roles": {
      "manager": "qwen3:4b",
      "executor": "qwen2.5-coder:32b",
      "verifier": "llama-3.3:70b"
    }
  }
}

If you're developing LocalClaw itself:

# Watch mode (auto-reload on changes)
npm run dev

# Test a single command without building
npx tsx src/cli/index.ts agent "test mission"

1. File operations: "Create 3 text files named file1.txt, file2.txt, file3.txt with different content"
2. Code generation: "Write a Python class called Calculator with methods for basic arithmetic"
3. Organization: "Create folders named src, tests, and docs in the workspace"
4. Processing: "Read all .txt files and create a summary.md file listing their names and sizes"

You know LocalClaw is working when:

• ✅ localclaw doctor shows all green checkmarks
• ✅ You can run localclaw agent "simple task" without errors
• ✅ Files appear in ~/localclaw/workspace/ after tasks
• ✅ localclaw jobs list shows your completed jobs

• Check logs: ~/.smallclaw/logs/
• Review database: ~/.smallclaw/jobs.db (SQLite)
• Enable verbose logging: Set environment variable DEBUG=*

Happy automating! 🦞