Use it
Safety Oracle API
A public REST endpoint that reads the verdict straight from the on-chain registry. No SDK, no API key, open CORS — the same call any agent makes. Base: https://tryanneal.xyz.
GET /api/safety/{codeHash}
Returns the on-chain is_this_safe verdict — SAFE or UNSAFE plus a 0–100 score — for a code hash. The default network is mantle-sepolia; pass ?network=mantle for the mainnet verdict (agent #131).
| Status | Meaning |
|---|---|
| 200 | Verdict found |
| 404 | No verdict on-chain for this hash |
| 400 | Malformed code hash |
| 502 | RPC failure |
The
safe flag is opinionated: any critical OR high finding flips it UNSAFE, even at a high score — a single critical at 90/100 still kills composability. Live proof: the Merchant Moe LB Router (~$60M TVL) reads back SAFE, 100/100, posted on-chain by agent #131.POST /api/safety/audit
Submit source for a live audit. Runs the full cascade when keys are configured, else falls back to Slither-only (mode: "static-only"). Rate-limited to 1 request / 5 min / IP.
Returns the verdict, findings, gas profile, corpus context, and the AES-GCM decryption key — once. TryAnneal never stores it.
Full spec, including the on-chain code hashes you can query today: safety API README.