Discover tools at runtime with AWS Agent Registry
Web page · https://aws.amazon.com/blogs/machine-learning/discover-tools-at-runtime-with-aws-agent-registry
Hardcoded tool wiring breaks down as an agent's catalog grows from a handful of tools to hundreds. This post walks through a runtime Discover, Instantiate, Execute loop built on AWS Agent Registry, where an orchestrator discovers MCP servers and A2A agents through semantic search at runtime instead of compiling every tool at build time.
First seen by Oriel on .
AI summary
The post demonstrates an orchestrator that uses AWS Agent Registry to find approved MCP servers and A2A agents at runtime, then builds a request-specific agent without hardcoded tool catalogs.
- The sample deploys an order-management MCP server through Amazon Bedrock AgentCore Gateway, two A2A agents on AgentCore runtime, a Registry, and an orchestrator.
- Hybrid keyword and semantic search returns record metadata and protocol descriptors, but invocation credentials must come from separate trusted configuration.
- Records require approval before they appear in discoverable search; publishers and approvers can have separate IAM-authorized permissions.
- For each request, the orchestrator creates MCP or A2A clients from search results, runs an ephemeral inner agent, and tears down MCP clients afterward.
- Three demos cover order status, pricing and discounts across MCP and A2A, and semantic matching for return and refund requests; the post cautions that larger catalogs do not inherently improve ranking.
#aws agent registry #mcp #a2a #agent orchestration
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| AWS Machine Learning Blog | 3a6f4d4b37bd5f432ae64ac9368553ce7e94b9ac | https://aws.amazon.com/blogs/machine-learning/discover-tools-at-runtime-with-aws-agent-registry/ Read discussion |