Building an MCP directory taught this dev that verification beats discovery
wagerx · reddit · 2026-08-28
A developer shares lessons from building an MCP/A2A service directory: finding endpoints is easy, verifying they actually work is the hard part. They found documented-but-dead endpoints, services that pass connection but fail tool calls, and unverifiable "MCP-compatible" claims.
The directory aims to answer: did it complete a real MCP handshake, which tools it exposes, whether tool calls were tested, when last checked, auth requirements, read-only vs. actions, and manual vs. automated monitoring. Key takeaway: discovery and verification should be separate — being listed shouldn't imply being verified. The author invites input on verification standards, check frequency, auth handling, and metadata design.
More from coding & agent
- Anthropic Guide: Building Reliable Claude Skills — bibryam · 2026-08-28
- Nirant: Claude and Codex Agent SDKs Don't Bend to Recursive Optimization — NirantK · 2026-08-28
- Grok Build Can Pull Up Claude Code Sessions — Works for Codex Too — kevinkern · 2026-08-28
- Deepseek Harness: Solving Agent Tool Contention via DB Monitoring — NirantK · 2026-08-28
- Deepseek Harness praised for god-tier cost-perf in iterative agent workloads — NirantK · 2026-08-28
- Forked Ninfer for TP2 to achieve 1M context with 50% throughput boost — Littlepharaoh · 2026-08-28