# Ask the Almanac

> Natural-language Q&A grounded strictly in the Almanac datasets, with citations.
> No generative model — every statement traces to a record. JSON: GET /api/ask?q=

## Use it

```
GET /api/ask?q=<your question>
```

Returns: `{ question, answer, grounded_in:[{type,id,title,json,page}], confidence, note }`

## Examples

- How much does Claude Opus 4.8 cost?  →  `/api/ask?q=How%20much%20does%20Claude%20Opus%204.8%20cost%3F`
- What is x402 and who created it?  →  `/api/ask?q=What%20is%20x402%20and%20who%20created%20it%3F`
- Does ClaudeBot honor robots.txt?  →  `/api/ask?q=Does%20ClaudeBot%20honor%20robots.txt%3F`
- What does Agent Experience mean?  →  `/api/ask?q=What%20does%20Agent%20Experience%20mean%3F`
- Which protocol connects agents to tools?  →  `/api/ask?q=Which%20protocol%20connects%20agents%20to%20tools%3F`

## How it works

Your question is tokenized and scored against every record in the Crawler Registry,
Protocol Atlas, Model Matrix and Lexicon; the best matches are composed into a
plain-language answer with citations. If the Almanac doesn't contain the answer,
it says so rather than inventing one. Grounding, not generation.