CompletionKit MCP: what works in Claude, ChatGPT, Cursor, and Grok Bot
Prompt evals over MCP: run a prompt on your dataset, score each output 1-5 with an LLM judge.
Registry name: com.completionkit/evals (v1.0.0) | GitHub stars
Status grid
| App | Status | Date tested |
|---|---|---|
| Claude | Untested | Not yet tested |
| ChatGPT | Untested | Not yet tested |
| Cursor | Untested | Not yet tested |
| Grok Bot | Untested | Not yet tested |
| MCP Inspector (reference client) | Untested | Not yet tested |
| MCP TypeScript SDK (reference client) | Untested | Not yet tested |
Hands-on test results
No hands-on tests yet. We do not publish results until we have actually tested.
Automated handshake
No public remote endpoint to check (local or package-only server).
Gotchas
None recorded yet. Gotchas come only from hands-on tests.
Compare with other tested connectors | Connector Weather
Request a hands-on test of CompletionKit
Get the badge
Maintain CompletionKit or write about it? Show its live ConnectorReport status in your README or docs.
Markdown (README, docs)
[](https://connectorreport.com/c/com-completionkit-evals?utm_source=badge&utm_medium=embed&utm_campaign=com-completionkit-evals)
HTML
<a href="https://connectorreport.com/c/com-completionkit-evals?utm_source=badge&utm_medium=embed&utm_campaign=com-completionkit-evals"><img src="https://connectorreport.com/badge/com-completionkit-evals.svg" alt="CompletionKit MCP status on ConnectorReport"/></a>
The badge updates on its own after every retest. How badges work