Plan
Convert intent into a canonical, typed Stellar ActionPlan.
Guardrails before execution. Turn AI and automation requests into typed Stellar actions, prove private-policy decisions, and verify them on-chain.
Evaluate
Apply deterministic allowlist, contract, amount, and approval rules.
Prove
Create a zero-knowledge receipt without revealing the owner policy.
Verify
Verify the known evaluator, decision, and audit binding on Soroban.
Automation can move fast. Authority should stay exact.
NeuroChain gives AI agents, bots, scripts, and backend jobs one deterministic boundary before a Stellar action can move forward.
Approved
The typed plan satisfies the committed policy.
Requires approval
The policy allows review, but execution remains stopped.
Blocked
Allowlist, contract-policy, or intent-safety rules reject the plan.
Running on testnet
Inspect the full proof path in the browser.
The hosted CLI exposes planning, all three decision states, read-only Soroban verification, and a separate explicit testnet attestation action.
Generate a wallet, run the ZK guardrail demo, verify the receipt on Stellar, and inspect the resulting testnet evidence without submitting the underlying ActionPlan.
Private where it matters. Public where it must verify.
The policy stays private. The proof publishes only the bindings needed to verify the evaluator and its decision.
payment != proof != approval != execution
Typed requestCanonical ActionPlan hash binds the exact requested Stellar action.
Private witnessOwner policy, commitment salt, and audit nonce remain off-chain.
Known evaluatorRISC Zero proves the deterministic NeuroChain guardrail program ran.
Public journalImage ID, hashes, policy version, decision, exit code, reason, and nullifier are bound together.
Soroban verifierStellar verifies the receipt and can reject invalid evidence or replay.
Use NeuroChain where automation already lives.
The same deterministic guardrails and private-policy ZK verification across every supported surface.
CLI and REPL
Inspect plans, guardrails, ZK status, and Stellar verification from one live developer surface.
Open Stellar CLIMCP and Skills
A no-submit tool contract gives agents a safe Plan, Evaluate, Prove, and Verify workflow with private-policy ZK proofs and read-only Stellar verification.
View the repositoryAPI and scripts
Use the same typed decisions from backend jobs, scheduled automation, and reproducible .nc flows.
x402 access
Paid access can sit in front of the service, but payment never bypasses policy or grants submit authority.
See the roadmapA valid proof is not permission to execute.
No-surprise-submit boundary
Verification confirms evidence. Approval and execution remain separate, explicit operations.