The minimum price per consultation is 10 satoshis, equivalent to a fraction of a cent on the dollar.
The service is designed for autonomous agents who pay alone, without human intervention.
An independent developer launched LLM402.ai on March 25, a service that allows you to consult artificial intelligence (AI) models by paying exclusively with bitcoin (BTC) through the Lightning Network (LN).
No account, email, or API key required: the only requirement is to pay a Lightning invoice for each query. According to the developer’s publication, the service is operational as of today.
Among the 32 models of AI available There are DeepSeek-R1, Llama (various versions), Qwen, Mistral and others, all accessible through endpoints compatible with the most used standards: the OpenAI and Ollama formats.
The price of each consultation It is dynamic and charged per request (pay-per-request). It is calculated mainly according to the chosen model, the number of input tokens (the text sent in the prompt and the context) and the maximum output tokens that are configured (max_tokens). That cost is then converted to satoshis using the current bitcoin price.
In this way, the dollar value tends to be quite stable, while the exact amount of satoshis varies depending on the BTC price at that time. The creator indicates that the cheapest consultations start at approximately 10 satoshis, equivalent to cents on the dollar today.
How does LLM402.ai work?
The LLM402.ai service uses the L402 protocol, a standard created by the company Lightning Labs that combines the code HTTP 402 (“payment required”) with Lightning invoices and authorization tokens called ‘macaroons‘, which certify payments in LLM402.ai.
The AI agent using LLM402.ai sends a request to this system, receives back a Lightning invoice along with a token macaroon, pay the invoice and resend the request including proof of payment.
With that test, the service delivers the model’s response. There is no prior registration or identity verification and the payment acts as the only credential.
The service is designed specifically for autonomous AI agents, that is, programs that operate alone without a human approving each action.
For that use, the creator indicates that LLM402.ai is compatible with lnget, a Lightning Labs tool that automates the entire L402 flow: detects the invoice, pays it, and resends the request without human intervention.
The service is also listed in the 402 Index, a directory launched this March 22 and which adds more than 15,000 endpoints of paid APIs for AI agentsmaking it easier for other agents to automatically discover you.
Bitcoin Lightning as an infrastructure for AI: the background
The launch of LLM402.ai comes at a time of convergence between AI agents and micropayments in Bitcoin Lightning, as CriptoNoticias has been reporting.
On January 31, the Lightning Enable team launched an open source tool that allows AI agents to make payments in Lightning via the Nostr Wallet Connect protocol, with programmable spending limits and payments restricted to BOLT11 invoices, the standard Lightning format that includes the recipient’s signature and fixed amount.
That tool is also built on L402 and targets the same model: agents who pay for digital services autonomouslywithout human intervention in each transaction.
What distinguishes LLM402.ai from the Lightning Enable kit and the Lightning Labs L402 protocol is that it is neither a tool for building agents nor a technical specification, but rather aservice through which an AI agent can query a language modelpay in satoshis and receive the response without a single line of additional configuration.
While Lightning Enable and Lightning Labs built the infrastructure, LLM402.ai uses it to deliver artificial intelligence as a commodity paid for by microtransaction. Each query is a complete Bitcoin transaction, settled in seconds, without a bank, without an account and without an intermediary.