Let's build a coding agent that can read files, write code, and run commands. Five minutes, start to finish.
1. Initialize
bash
npm install -g polpo-ai
mkdir my-agent && cd my-agent
polpo init2. Define your agent
Edit polpo.json:
json
{
"project": "my-coding-agent",
"teams": [{
"name": "dev",
"agents": [{
"name": "coder",
"model": "anthropic:claude-sonnet-4-20250514",
"systemPrompt": "You are a senior software engineer. Write clean, tested code.",
"tools": ["bash", "read", "write", "glob", "grep"]
}]
}]
}3. Test locally
bash
polpo start
# In another terminal:
curl http://localhost:1355/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model": "coder", "messages": [{"role": "user", "content": "Create a hello world Express server"}]}'4. Deploy
bash
polpo login
polpo deployThat's it. Your coding agent is live in production with sandboxed execution, persistent memory, and an OpenAI-compatible API.