High token use?

## Problem

According to [this comment](https://gist.github.com/amotl/e5f7bbcdb577a103fea6c49ed8afc100?permalink_comment_id=5582761#gistcomment-5582761), CrateDB's [llms-full.txt] weighs in rather heavy, with a token usage on OpenAI of:
```json
{"input": 212817, "output": 1139, "total": 213956}
```

## Solution

Instead of feeding the full uncompressed knowledge context, either serve individual elements on demand per MCP, or try to compress it.

- GH-6
- GH-30

[llms-full.txt]: https://cdn.crate.io/about/v1/llms-full.txt

Start by reviewing the linked llms-full.txt context and the discussion referenced in the issue, then read GH-6 and GH-30 for related approaches. Determine whether the project should serve individual knowledge elements through MCP or compress the full context; done means a documented, measurable reduction in token usage.

Written by the indexing model from the issue text.