DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

How to let your AI agent read documents through Docling with Docling

Connect Docling to an agent through its MCP server or run it as a service with docling-serve.

Short answer: Install Docling with pip install docling to test conversions locally first; then follow the MCP server docs to connect Docling to your agent; For a shared service, follow the API server docs for docling-serve. All steps.

  • Research
  • Intermediate
  • 30 minutes
  • From the README
  • Claude Code
  • Codex
  • Cursor

Who it is forSomeone using Claude Code or another MCP client who wants it to parse PDFs and office files properly.

The problemAgents often read PDFs as raw text and miss tables, figures and reading order.

The outcomeAn agent that can call Docling to parse documents, or a team service that converts files over an API.

What you need first

  • Python 3.10 or higher
  • macOS, Linux or Windows on x86_64 or arm64
  • Optional: a GPU for faster VLM pipelines such as GraniteDocling

Steps

  1. Install Docling with pip install docling to test conversions locally first
  2. Follow the MCP server docs to connect Docling to your agent
  3. For a shared service, follow the API server docs for docling-serve
  4. Ask the agent to parse a document and check the structure it returns

Prompt for your agent

Set up docling-project/docling for my agent using its MCP server, following the Docling MCP docs. Then parse <document> through it and summarise the tables it found.

Caveat. The README only links to the MCP and API server docs; setup details live there, not in the README.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for docling-project/docling and "Let your AI agent read documents through Docling"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer