Why Polpo?

Orchestration alone isn't enough. You need agents that deliver verified work.

AI agents are great at execution — terrible at finishing real work autonomously. Without oversight they drift, conflict, and stall. You end up babysitting the machines that were supposed to save you time: 4 monitors, 12 terminals, zero confidence anything actually works.

Polpo closes the loop. Describe what you need and walk away. Polpo assembles the right agents, plans the work, evaluates every result with LLM judges, retries what's broken, and escalates what it can't fix. It reaches you on Telegram, Slack, or email — only when it matters.

One npx command, no hosted platform, no vendor lock-in. Your laptop, your API keys, your rules.

How Polpo is different

| Approach | Coverage | What happens | |---|---|---| | Single agent | No coordination | One agent, one task. You babysit the chat and hope it finishes. | | Multi-agent (no review) | No quality checks | Multiple agents run in parallel — but nobody checks their work. | | Polpo | Full cycle | Orchestration, LLM-as-judge evaluation, automatic retry, and notifications. Work is done when it passes review. |

| | | | |:---:|---|---| | :robot: | Autonomous | Builds missions, picks agents, works through queued tasks 24/7. Walk away. | | :white_check_mark: | Reliable | Every task scored by LLM judges. Below threshold? The agent fixes it. You get results, not retries. | | :shield: | Crash-proof | Detached processes. Kill Polpo, reboot, lose connection — picks up where it left off. | | :bell: | Proactive | Reaches you on Slack, Telegram, email, or webhooks. You decide when and how. | | :repeat: | Playbooks | Define a mission once, run it forever. Schedule it, tweak it, improve it. Your agent teams get better over time. |

Installation

Zero-install (recommended)

npx polpo-ai

That's it. No global install needed. The setup wizard walks you through API keys, model selection, and team creation — then launches the dashboard.

Global install

npm install -g @polpo-ai/polpo    # or: pnpm add -g @polpo-ai/polpo

One-line installer (macOS / Linux / WSL)

curl -fsSL https://raw.githubusercontent.com/lumea-labs/polpo/main/install/install.sh | bash

Docker

docker run -it -p 3000:3000 -v $(pwd):/workspace lumea-labs/polpo

Quick Start

npx polpo-ai

The setup wizard detects your API keys, picks a model, and creates your first agent. Then you're in the dashboard — type what you need, hit Enter, and the agent does it.

Orchestrate a team

polpo mission create "Close $1M in revenue this month"

Polpo generates a mission with tasks, dependencies, and agent assignments. Agents execute, LLM judges score every result, and failures retry automatically.

Monitor (optional):

polpo status -w    # Live dashboard in another terminal
polpo serve        # HTTP API + Web UI at http://localhost:3000

Docker

docker pull ghcr.io/lumea-labs/polpo:latest
docker run -it -v $(pwd):/workspace ghcr.io/lumea-labs/polpo:latest

What your agents can do

Code. Browse the web. Send emails. Generate PDFs, Excel, Word docs. Create images, videos, audio. Search the internet. Talk to your customers. Handle credentials securely. And learn — Polpo creates and installs its own skills over time. Your agent teams get smarter the more you use it.

22+ LLM providers. 70+ tools. 5,000+ community skills. Free default model included.

Documentation

Full docs at docs.polpo.sh — configuration, API reference, guides, and more.

Contributing

We welcome contributions! See CONTRIBUTING.md for setup instructions, coding standards, and PR guidelines.

git clone https://github.com/lumea-labs/polpo.git
cd polpo
pnpm install
pnpm run build
pnpm run test -- --run

Security

If you discover a security vulnerability, please report it responsibly. See SECURITY.md for details.

License

MIT