54 MCP servers, 54 rated A on Glama, with tool definition scores from 4.0 to 4.8

The drawing

The problem

A directory grades an MCP server on how well its tools are described: the tool's name, what it does, each input, and its default. An agent picks a tool from that text alone. A vague definition gets the wrong tool called, or the right tool called with the wrong input, and nobody sees it happen.

What we built

Every server is generated from its actor's live input schema, read from the actor's current build, never from memory or a sibling server. Every field the actor requires is required in the tool. Every optional field carries the actor's own description with its default appended. Each server registers one read only tool, and its server key and package name follow the actor's slug. A committed test in each repo fails the build if a name copied from a sibling slips in.

On October 13 we read Glama's grading rule from its own release guide, its scoring post, and its public API, then read each listing's scores from the listing itself.

What it produced

• Domain to LinkedIn URL Resolver: score 4.5, and A on rating, quality, maintenance, and license after release 1.0.6.
• The quality grade comes from the tools a build exposes, so the score follows the definition text, not the listing copy.

Your scenario

If you publish an MCP server, read your tool text the way an agent does: the name, one sentence, and each input with its default. Then try one of ours, such as Domain to LinkedIn URL Resolver on Apify, from an agent you build. MindStudio builds AI agents without code. Send us your server's tool list and we'll tell you which definition we'd rewrite first.