Loading patterns…
Agent Registry & Discovery(ARD)
The discovery layer that lets agents and tool servers find each other at runtime instead of through hard-coded configuration. Three converging mechanisms make this work: A2A Agent Cards, a self-describing JSON manifest served at /.well-known/agent-card.json (RFC 8615) advertising an agent's skills, endpoints, auth schemes, and streaming capabilities; the official MCP Registry (preview September 2025), a federated public catalog and API for publishing and finding MCP servers that is extensible with private enterprise sub-registries; and decentralized naming proposals such as MIT Project NANDA, whose NANDA Index resolves cryptographically verifiable AgentFacts records to endpoints, positioned as DNS for agents. The common pattern is publish a signed capability record, then resolve it by well-known URI or registry query, validate the advertised auth schemes, and bind the agent dynamically. Distinct from `a2a-protocol`: that covers the message transport once a peer is known, while this covers finding and vetting the peer in the first place.
In 30 seconds
- What
- Agents publish signed capability records at well-known URIs or federated registries, then discover and validate peers dynamically instead of using hard-coded endpoints.
- When to use
- Orchestrators need runtime access to capabilities not built in, or ecosystems where agents and tools join and leave without redeployment.
- Watch out
- Registry staleness, compromised signing keys, or auth-scheme mismatches can wire in wrong or malicious agents before validation catches it.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Agent Registry & Discovery: Overview
The discovery layer that lets agents and tool servers find each other at runtime instead of through hard-coded configuration. Three converging mechanisms make this work: A2A Agent Cards, a self-describing JSON manifest served at /.well-known/agent-card.json (RFC 8615) advertising an agent's skills, endpoints, auth schemes, and streaming capabilities; the official MCP Registry (preview September 2025), a federated public catalog and API for publishing and finding MCP servers that is extensible with private enterprise sub-registries; and decentralized naming proposals such as MIT Project NANDA, whose NANDA Index resolves cryptographically verifiable AgentFacts records to endpoints, positioned as DNS for agents. The common pattern is publish a signed capability record, then resolve it by well-known URI or registry query, validate the advertised auth schemes, and bind the agent dynamically. Distinct from `a2a-protocol`: that covers the message transport once a peer is known, while this covers finding and vetting the peer in the first place.
- Agent Cards at /.well-known/agent-card.json (RFC 8615)
- Manifests advertise skills, endpoints, auth schemes, and streaming
- Federated MCP Registry: public catalog plus private sub-registries
- Signed capability records resolved by well-known URI or registry query
- Decentralized naming (NANDA Index, verifiable AgentFacts) as DNS for agents
- Auth-scheme validation before an agent is wired in dynamically
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
From the engineer behind this catalog
Get your agent architecture reviewed
This page documents one pattern. Your system runs dozens, and most failures live in how they fit together. Have the whole design reviewed against the 288 patterns in this catalog: architecture, reliability, evaluation and cost, every finding mapped to the pattern that fixes it.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September