AI inference for agents, paid in sats
Lampo is a public seller for pi402. Your Pi or OpenClaw agent buys model tokens over Lightning, per use: no account, no API key, no card, no chargebacks.
Start in one minute
pi install npm:@lightninglabs/pi-402
export L402_PROXY_URL=https://lampo.run/v1
export PI402_NETWORK=mainnet
export PI402_WALLET_PASSWORD='choose-your-own-strong-password'
export PI402_MAX_SAT_PER_REQUEST=5000 # Kimi and GLM bundles cost more than the 1,000 default
pi
/fund 20000 # pay the invoice from any Lightning wallet
/model pi402/deepseek-v4-flash # or pi402/glm-5.2, pi402/kimi-k3.1-pro, pi402/kimi-k3
On mainnet the pi402 wallet needs your own password and a seed you back up; see the pi-402 install guide. DeepSeek V4 Flash bundles fit pi402's default limits, so it works without raising PI402_MAX_SAT_PER_REQUEST.
Models and prices
One payment buys a bundle of tokens; each request is debited by the tokens it actually uses, input and output at their own rates. Prices follow the bitcoin price and include our 25% margin over the upstream provider.
| Model in pi402 | Bundle | Bundle price | Sats per 1M tokens |
|---|---|---|---|
| Loading current prices… | |||
Live counter
–bundles paid
–tokens served
–sats earned
–bundles in the last 24h
How it works
- Your agent sends a normal OpenAI-style request. Lampo answers
402 Payment Requiredwith a Lightning invoice for one bundle. - The pi402 wallet pays it within your spending limits and retries with the proof of payment.
- The response streams back. The same credential keeps working until the bundle is used up.
Good to know
- Answers are generated by AI models (Moonshot Kimi, Z.ai GLM, DeepSeek) and can be wrong. You are talking to software, not to a person.
- We do not store your prompts or responses: only token counts and payments.
- Payments are final. If something breaks on our side, write to us.
- pi402 and the Wavelength wallet are new software marked alpha by their authors. Keep small balances.