Handwriting OCR — Claude Code MCP server: what it does, how to install
Handwriting OCR runs handwriting recognition on scans and PDFs, then hands the result back to your agent as Markdown.
no license updated 2026-10-29
Open on GitHubWhat Handwriting OCR does
Three tools show up in your client: one uploads a document, one reports how the job is going, one returns the text. Images and PDFs go in; Markdown comes out.
The hosted endpoint at https://mcp.handwritingocr.com is the shorter route. Sign-in runs through OAuth in the browser, so nothing gets pasted into a config file. That version also exports to Word, Excel and JSON, and its extractors pull named fields off a page. Stdio-only clients reach it through the mcp-remote proxy.
Keep the local Node.js build for offline work. It needs Node.js 18.x or higher plus an API token exposed as API_TOKEN. The sample config points at someone else's absolute path to build/index.js — swap in yours, or the client will fail to launch it.
What it helps with
- Turn a photographed page into editable text
- Pull named fields out of a scanned form
- OCR a multi-page PDF without opening the web dashboard
- Hand the transcript to the assistant as Markdown
- See how far a long document has got in the queue
Example prompts for the agent
- Upload
receipt-scan.jpgto Handwriting OCR and show me the text. - Check the status of the document I sent and fetch the result.
- Export the finished transcription to Word next to the original file.
How to install Handwriting OCR
Install
The MCP registry has no launch command for it: see the repository README
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 Handwriting OCR
How do I install Handwriting OCR in Claude Code?
Point your MCP client at the hosted server by setting https://mcp.handwritingocr.com as the url value, then finish the OAuth sign-in. The local build is wired up by Smithery with npx -y @smithery/cli install @Handwriting-OCR/handwriting-ocr-mcp-server --client claude.
What does Handwriting OCR do?
It reads handwritten text from images and PDFs. Your assistant uploads a file, polls the job, and gets the transcription back as Markdown.
Do I need an API token?
Only for the self-hosted server, which reads API_TOKEN from the environment. The hosted endpoint authenticates through OAuth instead.