MCP PROMPT · AI KNOWLEDGE

/train-ai-agent

Add a knowledge document (URL or pasted text) so the merchant's AI agent can reference it.

01 · The flow

What the agent does, in order.

  1. 01 Confirm with the user: source type (URL or pasted text), title, where the content comes from.
  2. 02 If URL: call scrape_knowledge_url with the URL. The server fetches the body and stores it. Status will be "ready" or "failed" — if failed, suggest alternatives (try a different URL, or paste the text directly).
  3. 03 If pasted text: call add_knowledge_text with name + content. Use a descriptive name so the agent's retrieval system can disambiguate.
  4. 04 Optionally call list_knowledge_documents to confirm the doc landed and verify status.
  5. 05 Echo back the new documentId. Tell the user the AI agent will pick up the content on its next reindex (typically within minutes).
02 · Tools called

3 typed tools in sequence.

  • scrape_knowledge_url
  • add_knowledge_text
  • list_knowledge_documents

Every tool ships with MCP annotations (title, readOnlyHint, destructiveHint, openWorldHint, idempotentHint). Clients can gate destructive calls before invoking. Full catalog: /server-card.json.

03 · Related

More ai knowledge playbooks.

Try this in your agent.

Connect FavCRM to Cursor, Claude, ChatGPT, or any MCP client. Type /train-ai-agent in chat. Agent runs the steps in order.

Pick a client All 37 playbooks

Verified .