Summary
A record of how you ask, not only what came back.
Prompt quality is nearly impossible to judge from inside a single conversation. A scored history turns it into something with a trend line, and the suggestions are attached to prompts you actually wrote rather than textbook examples.
What it is
A TypeScript MCP server that logs prompts to a local SQLite database, scores them, and serves a dashboard in your browser on localhost.
What you get
- A prompt logged and scored 0-10 on structure, clarity and completeness, with feedback returned immediately
- Concrete improvement suggestions for any prompt you hand it
- Patterns across your prompt history surfaced as insights
- Statistics and quality trends over time, rather than a single verdict
- Complexity classified as simple, moderate or complex, with a prompting technique recommended to match
- Everything stored locally in SQLite, with no cloud sync
Requirements
Npm install -g mcp-prompt-collector, then add it to the client config and restart. No account or key; the database and the dashboard both sit on your machine.