Home / Trust Log
Trust Log
- What it does
- One public, append-only log of every signed result from LLPRO's trust tools, with inclusion and consistency proofs anyone can check.
- Who it is for
- Agents and their builders.
- Start in one step
- Use it on this page, or ask your agent to call
trust_log_headat llpro.ai/mcp. - What to expect
- Live: open today. Free. No sign-up.
New to this? Start here.
One public log of every signed result our trust tools give out. Each result is added as a fingerprint (SHA-256), never the result itself. A signed head fixes every entry in place, so anyone can check that a result was logged and that nothing was later changed or removed.
Right now
| Entries | 234 |
|---|---|
| Root hash | bc3cf7e3864f8828327d94bcc568f6a4c8116377bf9b234d761dce53d29309aa |
| Signed at | 2026-11-16T01:49:46Z |
| Signed by | llpro1:fvxbo2bbadlipsk4ho6whxck6e (Ed25519, key, listed in our key history) |
By tool
| Tool | Entries |
|---|---|
| agent-identity-check | 25 |
| countersign | 88 |
| dispute-record | 27 |
| run-records | 12 |
| signed-handoffs | 39 |
| spend-limits | 27 |
| transaction-audit | 12 |
| verified-attribution | 2 |
| witness | 2 |
Check it yourself
- Get the signed head:
/trust-log/v1/get-sth. Check its Ed25519 signature over the canonical JSON of the head withoutsignature(sorted keys, no spaces, UTF-8). - Find your result: for Countersign the fingerprint is the stamp's
entry_hash; for other tools it is the SHA-256 of the canonical JSON of{"observation": ..., "signature": ...}. Ask/trust-log/v1/find?product=<tool>&result_sha256=<hex>. - Get an inclusion proof:
/trust-log/v1/get-proof-by-hash?hash=<leaf hash>&tree_size=<size>, and check it against the signed root. - Keep the last head you saw. Later ask
/trust-log/v1/get-sth-consistency?first=<old>&second=<new>: the proof shows the new log still holds every old entry.
Hashing and proofs follow Certificate Transparency (RFC 9162): leaf = SHA-256(0x00 || entry JSON), node = SHA-256(0x01 || left || right). Answers are JSON with hex hashes, not the RFC 9162 binary format.
Entries: get-entries. Everything as one file: trust-log.json.
Agents: tools trust_log_head, trust_log_find, trust_log_proof, trust_log_consistency at llpro.ai/mcp.
What it does not show
The log shows that a result existed by the time of a signed head and was not changed or removed afterwards. It does not show that the result was right,
and it does not stop us from signing a result we never log; a monitor that keeps heads over time is what catches a log that forks or shrinks.
Today the log runs on our own server and the only monitors are ours; outside monitors are welcome (the endpoints above are all they need).
A monitor should take the log key from our signed key history (llpro-keys.json, tool trust-log), checked with the Offline verifier, not from the log itself. Shared Results: any logged answer can be shared as a page with the tool result_share and taken down with result_revoke.
A new signed head follows new entries within seconds, and an unchanged log is re-signed every hour.
Kept per entry: tool name, kind, a short reference, the signer key id, the fingerprint and the time logged. Never the result, names or addresses.
For agents: this page as Markdown · JSON · llms.txt · MCP endpoint llpro.ai/mcp · agent card