--- title: "Use Handwriting OCR inside Claude and other AI assistants" canonical: "https://www.handwritingocr.com/blog/handwriting-ocr-mcp-server" pubDate: "2026-07-22T00:00:00.000Z" description: "Connect Handwriting OCR to Claude, Cursor, or VS Code with our hosted MCP server. Transcribe scans, extract tables and fields, and export to Word or Excel from inside the conversation. OAuth sign-in, no API keys." author: "Sam Prentice" --- > **TL;DR:** [Handwriting OCR](/) now runs a hosted MCP server at `https://mcp.handwritingocr.com`. Add it as a connector in Claude, Cursor, or VS Code, sign in once with OAuth, and your assistant can transcribe scans, pull tables and fields out of forms, and export to Word, Excel, or JSON, all inside the conversation and all on your own account. No API keys. You can now use Handwriting OCR directly inside Claude, and inside any other AI assistant that speaks the Model Context Protocol (MCP), including Cursor and VS Code. Ask your assistant to transcribe a scanned letter, pull the line items out of an invoice, or turn a stack of handwritten notes into a Word document, and it does the work in the conversation using your Handwriting OCR account. No copying files between tabs, no API keys to manage, no glue code to write. ## What is MCP, and why it matters here MCP is an open standard that lets AI assistants securely connect to outside tools and services. Instead of pasting a scan into a chat and hoping the assistant's built-in vision reads the handwriting, the assistant hands the document to an OCR engine built specifically for handwriting, then works with the accurate text it gets back. That difference is the whole point. General-purpose assistants are impressive, but handwriting, especially cursive, mixed print-and-cursive, or rushed notes, is exactly where they give ground. Our engine processes thousands of real handwritten documents every week, and now your assistant can reach for it the moment a page needs reading. ## What you can do Once connected, your assistant gains a small, focused set of document tools: - **Submit a document** (PDF or image) for processing. - **Transcribe** it to plain text, the default action. - **Extract tables** into a spreadsheet, or **pull specific fields** with a custom extractor (for example, every date and total off a stack of invoices). Tables and custom extractors are available on the Pro and Business plans. - **Export** the result to Word, Excel, JSON, or CSV and get a download link. - **Check status and list** your documents, so the assistant can follow up on a job it started earlier. Because everything runs as you, the results land in your [dashboard](https://dashboard.handwritingocr.com/) as well, alongside anything you upload the usual way. ## How to connect The flow is the same in any MCP-capable assistant: 1. **Add the connector.** In your assistant's connector or MCP settings, add a custom connector pointing at `https://mcp.handwritingocr.com`. In Claude, that is Settings, then Connectors, then Add custom connector. 2. **Sign in.** Approve the browser sign-in and consent screen with your Handwriting OCR account. It uses OAuth, so there are no API keys to copy. 3. **Ask.** Tell your assistant what you need done with a document, and it uses the tools on your behalf. That is the entire setup. Do it once and the tools stay available in future conversations. ## Good to know - **It uses your existing credits.** Each processed page costs 1 credit, drawn from your balance, and free-trial credits work. Checking status, listing documents, and exporting an already-processed result are free. - **Your settings are respected.** Documents processed through the server follow your account's auto-deletion setting (7 days by default), the same as any other submission. - **Broad format support.** PDF, JPG, PNG, TIFF, HEIC, and more, up to 20 MB per document. - **Self-host if you prefer.** The server is open source at [github.com/Handwriting-OCR/handwriting-ocr-mcp-server](https://github.com/Handwriting-OCR/handwriting-ocr-mcp-server), so you can run your own instance. The hosted server is the fastest way to start. Our AI is built specifically for handwriting, including the cursive and messy notes that general assistants fumble. Connecting it as an MCP tool is now the fastest way to put it to work without ever leaving the chat. [Try Handwriting OCR free](https://dashboard.handwritingocr.com/register) with 5 credits, then add the connector and transcribe your first document from inside your assistant.