Agent access rails
Operating tools now sit behind handshake, contract, trust, quote, credit, and status rails.
The tools do the labor. The rails decide whether the work is supported, whether the scope is clear, whether the requesting agent is known, what the work costs, which credits are active, and where the result should return.
Handshake
Agent Handshake Rail
First contact, capability negotiation, public-safe data posture, and preflight readiness before quote or run.
Contract
Agent Run Contracts
Scope, limits, evidence requirements, outputs, metering policy, callback policy, and exception treatment.
Trust
Agent Trust Registry
Agent identity, operator contact, purpose, callback events, requested capabilities, and public-safe data posture.
Quote
Quote + Credit Rail
Estimated credits, evidence requirements, settlement direction, credit entitlement, status response, and run authorization.
Run
Agent Run Gateway
Metered task creation, Work Order routing, result package links, usage receipts, and callback payloads.
Ledger
Agent Task Ledger
Run state events, metering trail, result attestation, receipt links, and callback continuation records.
Recovery
Agent Exception Rail
Exception tickets, recovery requests, replay packages, and safe continuation paths for paused or partial runs.