Agentic Workflow

Agentic Workflow

Description: Chain multiple CalibreMCP tools in one flow for discovery, search, and action. Supports FastMCP 3.1 sampling for autonomous multi-step ebook processing, batch operations, and import pipelines.

Trigger Phrases

• "Automate [task] across my library"
• "Batch process all new books"
• "Set up an import pipeline for [folder]"
• "Find and fix [issue] in my library"
• "Process my Calibre library automatically"

Tools

• manage_libraries(operation="list") — List available Calibre libraries.
• manage_libraries(operation="switch", library_name="...") — Switch active library context.
• query_books(search="...", tags=[...], authors=[...], sort="...", unread=True/False) — Structured book queries.
• calibre_metadata_search(query="...", top_k=10) — Semantic metadata search.
• search_fulltext(query="...") — Full-text content search.
• manage_viewer(operation="open", book_id=...) — Open book in Calibre viewer.
• manage_metadata(operation="show"/"update", book_id=...) — Read or modify book metadata.
• manage_tags(operation="add"/"remove", book_id=..., tags=[...]) — Tag management.
• manage_analysis(operation="duplicates"/"health"/"reading_stats") — Library analysis tools.

Workflow

1. Discover: manage_libraries(operation="list") to identify the active library. Switch via manage_libraries(operation="switch", library_name="...") if multi-library setup.
2. Search: Combine query_books with calibre_metadata_search or search_fulltext depending on intent. Use conversational returns (each tool returns success, message, data dict) to decide next step.
3. Analyze: Run manage_analysis() to detect duplicates, health issues, or reading patterns before taking action.
4. Act: Apply changes via manage_metadata, manage_tags, or bulk operations. For books to read, chain manage_viewer(operation="open") as final step.
5. Verify: Optionally re-run analysis to confirm changes took effect.

Batch Pipeline Pattern

For import workflows: manage_libraries(operation="list") → manage_libraries(operation="switch", library_name="import") → query_books(sort="timestamp", limit=50) for newly added → manage_analysis(operation="health") to check quality → manage_metadata fixes for any issues → tag normalized batch.

Example

"Open a random unread science fiction book." → manage_libraries(operation="switch", library_name="main") → query_books(tags=["sci-fi"], unread=True, sort="random", limit=1) → extract book_id → manage_viewer(operation="open", book_id=...).