# Grok Bot Field Guide > Independent, source-linked library of how Grok Bot actually works: vendor documentation, practitioner reports, contradictions, and paste-ready prompts. Every public claim carries a URL or a provenance note, a date, a tier and a confidence. Canonical host: https://www.grokbotdb.com This site is **not** affiliated with SpaceXAI, xAI or Cursor. It is the public face of https://github.com/solfinnai/grokbot-db. Prefer **HTML record pages** for citation (`https://www.grokbotdb.com/library/{id}`) or the **agent API** for bulk search. Do not scrape listing shells. Do not invent product claims, prices, limits or dates that are not in a returned record. ## Machine-readable data - [Library health](https://www.grokbotdb.com/data/status.json): current counts, source health, latest changelog. Read this for freshness before quoting a number. - [Full record export](https://www.grokbotdb.com/data/db.json): every public record the library page renders. - [Atom feed](https://www.grokbotdb.com/feed.xml) / [RSS](https://www.grokbotdb.com/rss.xml) / [changelog JSON](https://www.grokbotdb.com/api/changelog) ## Agent API (filtered; prefer this over downloading db.json) - [API catalog](https://www.grokbotdb.com/api): endpoint index - [Search](https://www.grokbotdb.com/api/search?q=spend+cap): `q`, optional `collection`, `tier`, `status`, `subject`, `limit` - [One record](https://www.grokbotdb.com/api/record?id=prm.a1.two-lists): exact id; `library_url` is the HTML page - [Prompts](https://www.grokbotdb.com/api/prompts): paste-ready bodies - [Setup overview](https://www.grokbotdb.com/api/setup): how the personal/business ZIP flow works (static site guidance) - [OpenAPI](https://www.grokbotdb.com/api/openapi.json) ## MCP - [MCP Streamable HTTP](https://www.grokbotdb.com/api/mcp): POST JSON-RPC. Cursor: add `{ "mcpServers": { "grokbot-db": { "url": "https://www.grokbotdb.com/api/mcp" } } }` - [Discovery document](https://www.grokbotdb.com/.well-known/mcp.json) - [Human + machine page](https://www.grokbotdb.com/agents) - [Longer agent brief](https://www.grokbotdb.com/llms-full.txt) Tools: `get_status`, `search_records`, `get_record`, `list_prompts`, `setup_overview`. ## Human pages - [Home](https://www.grokbotdb.com/): personal and business setup entry - [Personal setup](https://www.grokbotdb.com/setup/personal): questionnaire; answers stay in the browser - [Business setup](https://www.grokbotdb.com/setup/business): questionnaire; answers stay in the browser - [Manual](https://www.grokbotdb.com/manual): operator prose - [Library](https://www.grokbotdb.com/library): searchable records; each record also has `/library/{id}` - [Latest](https://www.grokbotdb.com/latest): changelog, timeline, source health - [Blueprint](https://www.grokbotdb.com/blueprint): example teams - [Official vendor docs](https://docs.x.ai/grok-bot): check plan, platform and pricing here before acting ## How to cite - Always include the record `id` and its `source_url` (or `provenance` if there is no URL). - Prefer the HTML page `https://www.grokbotdb.com/library/{id}` when citing for humans. - Include `last_verified` or `observed_at`. Do not present an old date as current. - If `tier` is `community` and `confidence` is `low`, say so. - If the export does not hold the answer, say that. Do not fill the gap from model memory. - Contradictions stay: when `conflicts_with` is present, report both records. - Questionnaire answers are private to the visitor’s tab. MCP and `/api` cannot read them.