AnyDoc
Add local conversion from office documents to Markdown with the pinned Firecrawl AnyDoc CLI.
What it does
- Converts Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, text, and text-based PDF files
- Runs locally inside the agent container
- Does not upload documents to a hosted parser
- Pins the AnyDoc CLI and verifies its native binding
- Adds a focused document-conversion skill to agent containers
What you'll need
- NanoClaw installed and running
- Docker available to rebuild the shared agent image
Run the skill
Claude Code: /add-anydocCodex: $add-anydocOpenCode loads the same skill through its skill tool.
How it works
The /add-anydoc skill installs a pinned version of @firecrawl/anydoc in the shared agent image and adds a container skill that teaches agents how to convert supported documents to Markdown. The conversion happens inside the local Docker container. It does not add credentials, an MCP server, or a hosted parsing service.
The installer checks NanoClaw’s shared image manifest, copies the document-conversion skill, runs the dependency guards, rebuilds the image, and restarts affected groups. It also exercises the AnyDoc native binding instead of relying on a help command alone.
Limits
- Image-only PDFs need a separate OCR tool.
- Embedded visuals may not survive conversion completely.
- Spreadsheet Markdown is useful for reading, but it is not authoritative for calculations.
- Large or untrusted documents can consume substantial memory. Set a container memory limit for agents open to unknown senders.
Read the current source skill for the full install and verification flow.