Agent proposes
A structured intent — never free-form instructions or calldata hidden in extra fields.
SHADOW MODE · NO KEYS
Pre-sign policy firewall for AI agents
Autonomous agents.
Bounded authority.
ChainMandate gives agents a deterministic answer before an onchain action reaches a wallet: allow, deny, or request human review — with evidence another agent can verify.
No custody · No signing authority · No investment selection · No prompt execution
DETERMINISTICStructured inputs, decimal-safe policies
FAIL CLOSEDUnknown means deny or review
AGENT NATIVEJSON discovery, OpenAPI, verifiable receipts
Network reference layer
See the system.
Before it is live.
Explore how a future operator network could turn agent jobs into committee work, commit/reveal quorum and exact fee allocation—without pretending a testnet, token or settlement already exists.
01
Network explorer
Simulate Job → Committee → Commit → Reveal → Quorum → Fee Split.
OPEN REFERENCE ↗02Operator workbench
Inspect roles and download a closed application JSON locally.
PREPARE FILE ↗03Mobile companion
Inspect discovery and readiness from iPhone, iPad or Android; pairing and control remain off.
VIEW PRODUCT ↗04Control cockpit
Model traffic, price and mix while tracking honest launch states.
VIEW PLAN ↗NETWORK: PUBLIC REFERENCE ONLY / NOT LIVE · CORE API X402: CODE READY / NOT DEPLOYED · TOKEN: NOT ISSUED · VERIFIED USERS/REVENUE: NOT REPORTED
Machine first
Discover. Evaluate.
Verify.
An agent never needs to parse this page. The catalog exposes capabilities, boundaries, pricing status and callable endpoints as JSON. The free demo accepts only structured data.
01 SERVICE DISCOVERY
GET /.well-known/chainmandate.json
{
"audience": ["ai_agent", "agent_platform"],
"mode": "advisory_shadow",
"execution": false,
"mcp": "separate_adapter_not_deployed",
"x402": "core_api_implemented_not_configured"
}SigFence evaluates
An immutable mandate plus connector-attested market, balance and position state.
Agent verifies
A short-lived signed receipt proves exactly which intent and policy produced the answer.
What the mandate understands
A wallet limits access.
A mandate limits intent.
Mandate checks
Venue, chain, instrument, notional, leverage, exposure, loss, rate, slippage and stop rules.
Verified state
Caller claims never create an allow. Account and market state must be connector-attested.
Signed evidence
Each decision is bound to intent, policy version, wallet, control epoch and expiry.
Emergency control
A kill switch revokes prior allows. Only a proven risk-reducing close may pass.
No-key sandbox
Test the decision
contract.
This public endpoint demonstrates response semantics only. It never receives a wallet, private key, transaction, real position or production approval authority.
WAITING FOR INTENT—
Change the structured inputs and run the policy.
Planned launch pricing
Pay for control.
Never for trades.
These are non-binding launch assumptions for future software access — never fees on assets, volume, fills, profit or performance.
DISCOVERY
€0forever- Machine-readable catalog
- Public decision sandbox
- No production receipts
PLANNED AGENT PILOT
BUILDER
€249/ month- 3 registered agents
- 10,000 shadow evaluations
- 30-day decision audit
PRODUCTION
€1,490/ month- 25 registered agents
- 100,000 evaluations
- 12-month audit retention
REVALIDATE0.01 USDCreceipt check
POLICY0.05 USDCsingle decision
SIMULATE0.15 USDCtransaction model
QUORUM0.25 USDCcommittee decision
Q-SIMULATE0.75 USDCcommittee model
Planned agent-native PAYG rates are model assumptions, not an offer. The core API has a fail-closed x402-v2 payment and reconciliation path, but it is not configured or deployed. No paid plan, contract or agent charge is activated by this Site.
The mandate comes before the signature