Receipts for what your agents do.

When an autonomous agent takes an action — sends a payment, approves a refund, calls a tool, writes a record — Agent Receipts mints a signed, tamper-evident proof it happened, with what inputs, what outputs, and when. Verifiable by anyone, forever.

ed25519 signedSHA-256 content hashhash-chained per agent x402 pay-per-callMCP-nativekeyless USDC on Base

Read the developer docs → Run the live demo →

Why this exists

Audit trail for agentic systems

Agents act faster than humans can review. A receipt is the durable, after-the-fact record of every consequential action.

Compliance for AI actions

Prove to an auditor or regulator that a specific agent took a specific action on specific inputs at a specific time.

Dispute resolution

"Did the agent really approve this?" The content hash + signature settle it: tamper changes the id, breaking the proof.

Tamper-evidence

Each receipt commits the hash of the agent's previous receipt. Delete or alter one and the chain breaks — provably.

How it works

1. issue_receipt: Agent (or operator) submits action, agent_id, and inputs/outputs (we hash them; raw data is never stored).
2. we sign: We canonicalize + SHA-256 the receipt body, derive a content-addressed id, ed25519-sign it, and link it into the agent's hash chain.
3. verify_receipt: Anyone re-hashes the body, checks the id commitment, and verifies the signature. Free. No login.

Issuer signing key

Algorithm: Ed25519
Public key: 9107b9ca657ed10c424491ed45b8af487e9a4abb8318627ada9f6110773fb2bc
Receipts issued: 8

Verify any receipt's signature against this key offline. Machine-readable at /pubkey.

Agent Receipts · verifiable provenance for autonomous-agent actions · ed25519-signed, SHA-256 content-hashed, hash-chained.
Home · Developers · MCP · /.well-known/x402 · /llms.txt · issuer key