AiFinPay — Payment Rail for AI Agents · AIFP-1 x402 payment protocol

Per-call on-chain payment rail for autonomous AI agents. Live across 13 mainnet networks — Polygon is the SDK default, Solana is native; Avalanche, BNB Chain, Arbitrum, Base, Unichain and more settle through the same atomic fee-on-top split (98.99% merchant / 1% protocol treasury / 0.01% creator). Agents pick the chain they natively hold funds on — same SDK call, same fee model everywhere. One SDK call (agent.call({provider})) → one HTTP 402 challenge → one atomic on-chain settlement → the upstream service response. Non-custodial by design: every agent generates and holds its own wallet; AiFinPay never holds keys or funds.

For MCP-aware agents (Claude Code, Cursor, Continue, Cline)

If you are an LLM agent running inside an MCP host, your operator installs the AiFinPay MCP server with one command and restarts the host: 

claude mcp add aifinpay -- npx -y @aifinpay/mcp

After restart the host exposes 16 MCP tools — 7 payment tools (agent_address, agent_call, agent_claim_self, payable_fetch, agent_quote, pay_with_split, quote_split) plus 9 read-only directory tools (list_providers, network_stats, leaderboard and more). The agent generates and controls its own wallet — keys are created locally and never leave the agent's environment. Full setup, wallet top-up and dashboard-claim instructions live in the agent playbook: aifinpay.io/llms.txt · npm package.

For non-MCP agents using the SDK directly

npm install @aifinpay/agent (or pip install aifinpay-agent), then follow the step-by-step integration guide in /llms.txt and the open-source examples at github.com/AiFinPay/sdk. Wallets are generated locally by the SDK and stay under the agent's sole control (non-custodial).

Live providers

Always-current list: aifinpay.io/api/registry.

  • Exa AI search — neural/fast/deep modes — native x402 on Base USDC + bridged on Polygon. Slug: exa.
  • Venice AI — open-source LLM inference on GPU compute, OpenAI-compatible API. Slug: venice.
  • io.net IO Intelligence — managed LLM inference on io.net's decentralised GPU pool (Llama 3.3, Mixtral, more). Slug: io-net. First live x402 → io.net call: 0x7c6ca0ffcf75b1ca3ade4800fb896c4bb08bc5f1a91916dc2cf4918f16129f0a.

API endpoints

  • GET https://aifinpay.io/llms.txt — full agent integration playbook (SDK + MCP + dashboard claim).
  • GET https://api.aifinpay.io/registry — full provider catalog with prices, modes, accepted chains.
  • GET https://api.aifinpay.io/registry/coverage — category coverage (live + recruiting).
  • GET https://api.aifinpay.io/.well-known/x402.json — x402 discovery document.
  • GET https://api.aifinpay.io/manifesto.json — protocol manifest.
  • POST https://bridge.aifinpay.io/{slug}/... — per-service x402 paid proxy (returns 402 then 200 after on-chain settlement).
  • GET https://api.aifinpay.io/nonce — Solana Ed25519 nonce (legacy gate, kept for the Seat-PDA identity layer).

Settlement contract

Polygon mainnet AiFinPay Splitter: 0xE34Fc0E6694821c600Fa0955C0F74720ea6d8440. Atomic split per paid call: 98.99% to the merchant wallet, 1% protocol fee to the AiFinPay treasury Gnosis Safe 0xD31d82c4b35DABaA2ad7023C89A78A052D1f3c8e, 0.01% optional creator pointer. Function signature: payMatic(address merchant, address ipCreator, string orderId). These addresses are published for independent on-chain verification only — the SDK settles through the contract; nothing is ever sent to them manually.

Loyalty layer (mSECCO)

mSECCO is an internal accounting unit — non-transferable usage credits, accrued automatically as agents pay for calls through AiFinPay. Not a token, not a speculative asset.

Operator

AiFinPay — AI Agent Payment Protocol. Governance: 3-of-4 Squads Multisig (Solana), Gnosis Safe (Polygon). Non-custodial throughout: AiFinPay never holds agent private keys or upstream service API keys.