ReadAnyDoc — Claude Code MCP server: what it does, how to install
ReadAnyDoc — an MCP server for Claude Code that converts PDFs, scans, and Office files into Markdown or typed schema fields with bounding boxes. Runs document pages through vision models.
Apache-2.0 ★ 11 updated 2026-11-16
claude mcp add readanydoc -e ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY> -- uvx readanydocWhat ReadAnyDoc does
ReadAnyDoc processes document pages using vision models (Claude Haiku 5.5 by default) and extracts tables, Markdown text, or schema fields with exact bounding boxes for every value. Its layout engine identifies block types, forms, headers, and footers.
Two execution modes are supported: full engine mode using local CPU layout models, or a minimal model-only mode. On its first run, the engine downloads roughly 200 MB of model weights, and optionally relies on system Tesseract.
Operation requires Python 3.12 and an active ANTHROPIC_API_KEY or OPENAI_API_KEY.
What it helps with
- convert scanned documents and multi-page PDFs into clean Markdown
- extract invoice fields into structured schemas with exact bounding boxes
- process handwritten document pages using Claude Sonnet
- preserve complex table layouts without manual block cropping
Example prompts for the agent
- Parse
invoice.pdfand convert it into Markdown format - Extract total amount and date from
statement.pdfwith citations - Process scanned document
form.pdfand extract all table blocks
How to install ReadAnyDoc
Platform CLI
claude mcp add readanydoc -e ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY> -- uvx readanydocReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from readanydoc <command>
Platform CLI
codex mcp add readanydoc --env ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY> -- uvx readanydocReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from readanydoc <command>
Config file
{
"mcpServers": {
"readanydoc": {
"command": "uvx",
"args": [
"readanydoc"
],
"env": {
"ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from readanydoc <command>
Config file
{
"servers": {
"readanydoc": {
"type": "stdio",
"command": "uvx",
"args": [
"readanydoc"
],
"env": {
"ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from readanydoc <command>
Platform CLI
gemini mcp add -e ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY> readanydoc uvx readanydocReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from readanydoc <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about ReadAnyDoc
How to install ReadAnyDoc in Claude Code?
Run claude mcp add readanydoc -e ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY> -- uvx readanydoc.
What does ReadAnyDoc do?
It converts PDFs, images, and Office documents into Markdown, tables, or typed schemas while returning exact bounding box coordinates for each extracted field.
What is required to run ReadAnyDoc?
You need Python 3.12, uvx, and an API key such as ANTHROPIC_API_KEY or OPENAI_API_KEY. The engine downloads approximately 200 MB of weights on first run and can utilize system Tesseract.