Frontier models, paid per answer.
Your agent asks a question, gets an answer from GPT-6.1, Claude, Gemini or Grok, and pays for that one answer. No provider accounts, no API keys, no monthly bills: one wallet buys every model.
What it is good for
Any step you would send to a chat API
Drafting, summarising, extracting, classifying, deciding. Tool calls and JSON output work. Answers arrive whole, not streamed.
Pick the model per call
A cheap, fast model for routine steps and a frontier model for the hard ones, in the same place, from the same wallet.
Quoted before it runs
The price comes from your request size and max_tokens. Typically a few tenths of a cent for a short exchange, a few cents for a long one. A failed call is free.
Nine models, one price list.
Example prices for a call with a thousand tokens in and a thousand out. Thinking models are on by default at low effort.
| Model | Tier | Example: 1k tokens in, 1k out |
|---|---|---|
Claude Haiku 5.5anthropic/claude-haiku-5.5 | fast | $0.002 |
Gemini 3.8 Flashgoogle/gemini-3.8-flash | fast | $0.0068 |
GPT-5.4 miniopenai/gpt-5.4-mini | fast | $0.0079 |
Grok 4.7xai/grok-4.7 | frontier | $0.0117 |
Claude Sonnet 5.5anthropic/claude-sonnet-5.5 | frontier | $0.0173 |
GPT-5.5openai/gpt-5.5 | frontier | $0.0495 |
GPT-6 Lunaopenai/gpt-6-luna | fast | $0.002 |
GPT-6.1 Terraopenai/gpt-6.1-terra | frontier | $0.0089 |
GPT-6.1 Solopenai/gpt-6.1-sol | frontier | $0.0173 |
Price per call = your request's tokens at the provider's list rate, times 1.4, plus half a cent; never under $0.002. Current rates, machine-readable: /ai/v1/models. More on the pricing page.
Two ways to call it.
An OpenAI-compatible endpoint that any x402 HTTP client can pay, including Coinbase's wallet from Claude.ai, ChatGPT or Claude Code; or, for agents you build with a paying MCP client, an MCP server with a chat tool.
Plain HTTP
Unpaid, this returns 402 with the quote. An x402 client (@x402/fetch, the Python or Go SDKs, or the Coinbase wallet CLI) pays and retries, and the reply is an OpenAI-style completion.
curl https://payperwork.app/ai/v1/chat/completions -H 'content-type: application/json' \
-d '{"model":"anthropic/claude-haiku-5.5","messages":[{"role":"user","content":"hi"}],"max_tokens":256}'MCP, for agents you build
The client must be wrapped with an x402 payment layer (@x402/mcp or the agents package). Off-the-shelf clients connect but cannot pay; an unpaid call returns the price and the HTTP route above.
{ "mcpServers": { "payperwork-ai": { "type": "streamable-http", "url": "https://payperwork.app/ai/mcp" } } }The payer needs USDC on Base and no gas. Full route details in openapi.json and llms.txt.
What holds for every inference call.
Guaranteed
- The price is quoted from your request size and max_tokens before the model runs, and that quote is what you pay.
- If the model call fails you are not charged.
- Payperwork does not store prompts or completions.
Not guaranteed
- Privacy at the provider: prompts go to the model's provider (Anthropic, OpenAI, Google or xAI) and the provider's own retention applies.
- Output budget on thinking models: reasoning tokens count against max_tokens and are billed as output; give such models room or set reasoning_effort.
- Streaming: responses arrive whole.
Give your agent every frontier model.
One wallet, no provider accounts, priced per answer.