HackerTrans
トップ新着トレンドコメント過去質問紹介求人

warmcat

no profile record

投稿

[untitled]

1 ポイント·投稿者 warmcat·4 か月前·0 コメント

Ask HN: With Promptfoo acquired by OpenAI, what are MCP devs using for testing?

2 ポイント·投稿者 warmcat·4 か月前·4 コメント

Show HN: MCPSpec – Ship reliable MCP servers without writing test code

light-handle.github.io
7 ポイント·投稿者 warmcat·5 か月前·2 コメント

コメント

warmcat
·4 か月前·議論
Plus 1, Goodnotes is such a well developed product. I also have an iPad screen guard which gives a paper like feel when writing which makes my brain think I am writing on a book.
warmcat
·4 か月前·議論
Good work. But what makes this different than just another feature in Gemini Code assist or Github copilot?
warmcat
·4 か月前·議論
This is a good list. Are you finding these can be integrated into your CI/CD workflows?
warmcat
·5 か月前·議論
There are a few tools in the space, but they each focus on parts:

- MCP Inspector (Anthropic's official tool) — interactive debugging, not CI-oriented - MCP-Scan (Invariant Labs, now Snyk) — security scanning, focused on tool poisoning detection - Promptfoo — LLM red teaming tool that added MCP support recently - MCP Protocol Validator — spec compliance checking

MCPSpec tries to be the one tool that covers the full workflow: record, replay, mock, security audit, quality scoring, and CI setup. None of the above do recording/replay or mock generation.

As for standards — there aren't any yet. MCP itself moved under the Linux Foundation's Agentic AI Foundation in December 2025, and NIST launched an AI Agent Standards Initiative last month. But no conformance or testing standard exists. It's still early.