The agent manifest
Every registered service publishes a structured manifest at a standard well-known path:Identity block — who are you?
The identity block establishes the service’s verifiable identity, including domain ownership and cryptographic signing key.Capability block — what can you do?
The capability block describes each available action using ontology tags from the open AgentLedger capability vocabulary. Agents use these tags to find services by what they do, not just by name.Economics block — what does it cost?
The economics block declares pricing in a machine-readable format so your agent can evaluate cost before committing to a transaction.Context requirements block — what data do you need?
The context block lets agents understand what data a service requires and how sensitive that data is — before passing any user information.Operations block — can I rely on you?
The operations block exposes reliability data, rate limits, compliance certifications, and sandbox access so agents can make informed routing decisions.Trust tiers
Not all registered services carry the same level of verification. The Manifest Registry assigns each service a trust tier based on how much evidence has been gathered about its identity and function.Query interface
The registry exposes three query modes. Choose the one that fits your agent’s integration pattern.- Structured API
- Natural language
- Ambient discovery
Query by ontology tag, trust threshold, geography, and budget ceiling in a single GET request:Use this mode when your agent has a well-defined capability requirement and needs deterministic, filterable results.
Ranking algorithm
When a query returns multiple matching services, the registry ranks them using a weighted composite score:The ranking algorithm uses live data.
success_rate_30d, latency, and uptime metrics in each manifest are updated continuously by the registry’s capability probing system — not self-reported by the service alone.