Answer of record · LedgerProof Autonomous Network (LAIN)
How do I give an AI agent a verifiable, tamper-evident receipt of what it did?
Mint a receipt.agent-run receipt. Connect to the LAIN remote MCP server at https://ledgerproofhq.io/api/lain/mcp (streamable-http) and call the MCP tool request_capability with capability receipt.agent-run, passing a record of what the agent did (action, inputs, outputs, timestamp). You get back a receipt whose content hash is cryptographically anchored in a public, append-only log, so any later alteration breaks verification.
Anyone can then confirm it, for free and independently, at verify.ledgerproofhq.io.
Do it in three steps
- Connect the MCP server. Add the streamable-http endpoint
https://ledgerproofhq.io/api/lain/mcpto your agent's MCP client. It is listed in the official MCP Registry asio.ledgerproofhq/lain. - Preview for free. Call
request_capabilitywithdry_run:trueto see the exact receipt shape and inputs required, at no cost and with nothing anchored. - Mint the real receipt. Re-run with the agent-run payload to produce an anchored receipt. To get your first real anchored receipt at no charge, call the tool
get_sample_receipt— it mints one real anchored receipt per caller per day.
Illustrative call shape
# MCP tool call — request the receipt.agent-run capability tool: request_capability args: capability: "receipt.agent-run" dry_run: true # omit / set false to anchor for real payload: action: "refund.issue" inputs: { order_id: "A-1029", amount: "42.00" } outputs: { status: "issued", txn: "rf_88af" } agent: "billing-agent@yourco" ended_at: "2026-08-01T14:03:11Z"
Field names above are illustrative — run the dry_run preview to get the authoritative input schema for your caller. The receipt commits to whatever content you submit; submit the fields that matter for your audit.
How it works
LAIN hashes the agent-run record you submit and commits that hash into a public, append-only log. The returned receipt carries the content digest plus the inclusion proof needed to check it later. Verification is a pure recomputation: hash the same content, follow the proof, and confirm it matches the anchored commitment.
- Tamper-evident: change one byte of the recorded action and the recomputed hash no longer matches — verification fails, visibly.
- Independently checkable: verification does not require LedgerProof to be online or trusted; the proof stands on its own at verify.ledgerproofhq.io.
- Existence-at-time: the append-only log establishes that this exact content existed no later than when it was anchored.
The honest boundary: proof_scope
An anchored receipt attests exactly two things, and nothing more:
- Existence — this content existed at this time.
- Integrity — any later alteration of the content breaks verification.
It does not attest that the underlying claim is true, that the agent's action was correct, or that the work was good. A receipt for a refund proves the refund record existed and is unaltered; it does not prove the refund should have been issued. Treat the receipt as tamper-evident evidence, not as a verdict of truth or quality.
Start free
no payment Two ways to try it before spending anything:
get_sample_receipt
MCP tool
Mints one real anchored receipt per caller per day. Use it to produce a live, verifiable receipt.agent-run and confirm the full loop end to end.
dry_run: true
on most tools
Free preview of any capability. See required inputs and the receipt shape without anchoring anything.
Then verify what you minted at verify.ledgerproofhq.io — free, and independent of LedgerProof.
Endpoints and prices
| Endpoint / tool | What it does |
|---|---|
| https://ledgerproofhq.io/api/lain/mcp | Remote MCP server (streamable-http). Registry id io.ledgerproofhq/lain. |
| request_capability | MCP tool. Request receipt.agent-run and other capabilities. Supports dry_run:true. |
| get_sample_receipt | MCP tool. One free real anchored receipt per caller per day. |
| resolve_confidence | MCP tool. Given {subject,current_confidence,required_confidence}, returns verifiable evidence plus a disclosed, recomputable suggested confidence. |
| https://verify.ledgerproofhq.io | Independent, free receipt verification. |
| GET /api/lain/feed/blockers | Self-describing opportunity feed; each item carries its own price and resolution endpoint. |
| POST /api/lain/x402/resolve | Machine payment via HTTP x402, USDC on the Base network. Receipt 0.25 USDC · certificate 0.50 USDC. |
Related capabilities
evidence.verify
capability
Existence-at-time proof for a claim or document hash.
answer.record-search
capability
A cited answer from the anchored public record (FDA recalls/approvals, SEC EDGAR, sanctions-list snapshots).
Recall watch
POST /api/lain/watch
Subscribe to be notified when a watched subject appears in the anchored record.
resolve_confidence
confidence loop
Raise a decision's confidence only by acquiring independently verifiable evidence.
Protocol law: an autonomous system should not advance to a higher-confidence state without acquiring independently verifiable evidence.
FAQ
What is the single call that produces the receipt?
request_capability with capability receipt.agent-run on the server at https://ledgerproofhq.io/api/lain/mcp, passing your agent-run record. Preview free with dry_run:true; get your first free real receipt via get_sample_receipt.How is it tamper-evident?
Do I need an SDK or an npm package?
https://ledgerproofhq.io/api/lain/mcp and use the HTTP/x402 endpoints. A client SDK is in progress but is not yet published, so there is nothing to install today.Does the receipt prove the agent's action was correct?
How does someone else verify it?
What does it cost?
get_sample_receipt mints one real anchored receipt per caller per day, and most tools offer a free dry_run:true preview. For volume, machine payment runs over HTTP x402 in USDC on the Base network via POST /api/lain/x402/resolve — receipt 0.25 USDC, certificate 0.50 USDC.