Astro Astro Open Safety Layer
A non-custodial boundary for fingerprinting, simulating and assessing blockchain intents before authorization, then issuing a verifiable safety receipt.
One check between intent and execution
Wallets, dapps, agents and bridges can use the same machine-readable decision without handing Astro control of user assets.
Normalize actor, actions, constraints, nonce and deadline.
Generate a deterministic SHA-256 identity for the complete intent.
Run static policy checks and, where configured, optional allowlisted RPC calls.
Return explainable signals and allow, review or block.
Sign the result using a published Ed25519 verification key.
Safety decisions with evidence
The reference engine checks what it can prove and labels what it cannot. Unknown never silently becomes safe.
- DestinationsInvalid EVM and zero-address detection.
- PermissionsERC-20 unlimited approval and operator-wide NFT approval detection.
- ConstraintsDeadline, aggregate value and human approval thresholds.
- SimulationThe adapter supports optional
eth_calland gas estimation through configured RPC endpoints. The hosted beta currently reportsrpcChains: [], so live RPC simulation is unavailable there. - Threat feedEvidence submissions remain pending until reviewed.
POST /v1/check { "intent": { "version": "aoi.intent/0.1", "actor": { "type": "agent" }, "actions": [ ... ], "constraints": { "requireHumanApprovalAbove": "20" } }, "liveRpc": false } decision: "allow" | "review" | "block" signals: explainable evidence[]
Small surface, portable integration
The TypeScript reference implementation includes OpenAPI discovery, an SDK, Docker deployment and request controls. Inspect the hosted beta through its protocol discovery document or OpenAPI specification.
| Method | Endpoint | Purpose |
|---|---|---|
| POST | /v1/intents/fingerprint | Validate and fingerprint an AOI intent envelope. |
| POST | /v1/simulate | Run deterministic checks; live chain simulation requires a configured RPC endpoint. |
| POST | /v1/check | Return simulation plus an explainable safety decision. |
| POST | /v1/route | Rank integrator-supplied routes by fee, speed and safety. |
| POST | /v1/receipts | Issue an Ed25519-signed safety receipt. |
| POST | /v1/receipts/verify | Verify receipt integrity against the published key. |
| GET | /v1/safety/{chain}/{subject} | Query the open safety feed without assuming unknown is safe. |
Implemented claims stay separate from rollout plans
Status checked 27 September 2026: the hosted beta health response reports rpcChains: []. Static checks and signed receipts are available; live RPC simulation remains unavailable until allowlisted chain endpoints are configured.
Implemented and tested
- Public beta endpoint with persistent Ed25519 signing key
- Intent schema and canonical fingerprints
- Static security rules and explainable scoring
- Allowlisted EVM RPC simulation adapter in the reference implementation
- Route candidate ranking
- Signed receipt issue and verification
- Safety-feed lookup and moderated submission
- TypeScript SDK, Docker and automated tests
Next public rollout
- Configure and validate allowlisted hosted RPC endpoints
- Custom API domain, key history and uptime dashboard
- Append-only receipt storage
- Reviewed threat-intelligence dataset
- Independent security audit
- Wallet and dapp compatibility pilots
- Additional chain-specific simulation adapters