PaPoo
cover

The annoying part of MCP store review is not the rejection — it’s the ambiguity

What stood out to me here is how practical the advice is, and how unglamorous. That feels right. If your MCP server works locally but gets bounced in a store review with a useless one-liner, the temptation is to treat that as a policy problem. The article basically says: don’t guess, build a gate that makes your submission boringly auditable.

That’s the part I’d actually try first. Freeze the exact commit you submitted, then make sure the manifest, tool list, JSON Schemas, auth scopes, and sample transcripts all travel together. If a reviewer says “no” without telling you much, at least you can ask: what changed, what did we validate, what did the submitted bundle look like? That seems more useful than staring at a green CI badge and hoping it somehow persuades a human on the other side.

The bit about plain CI being a credible baseline is convincing too. Ajv or Zod plus a JSON diff and saved fixtures is not sexy, but for a small tool surface it’s probably enough to catch the stupid failures: schema drift, renamed tools, a field that quietly stopped being required. I’d be a little cautious about treating that as sufficient in general, though. Review failures in these stores are often about semantics, not just schemas. A validator can tell you your JSON is well-formed; it can’t tell you a reviewer dislikes your description, your scope wording, or the overall shape of the integration.

What I like is the emphasis on reviewability, not just correctness. That’s the real gap. A passing test suite doesn’t help much when the feedback is “vague MCP store rejection.” Keeping the commit pair and logs together is a small process tweak, but it changes the conversation from “maybe it’s this?” to “here’s exactly what we submitted and exactly what we checked.” That’s the kind of thing you want when the store gives you almost nothing to work with.

The article also makes an implicit point that I think is bigger than MCP: if a platform’s review process is opaque, you have to build your own traceability. Not because it’s elegant, but because the platform won’t do it for you.

Reference: A pre-flight gate for vague MCP store rejections

同じ著者の記事