I created Go CLI tool for processing documents with Mistral AI’s OCR with the help of Claude using MCP. Yesterday, I saw an announcement from Mistral about their OCR. I wanted to try it and check if it works for Burmese text. In short, it does not work. It give me Thai text instead.

The following is the original Burmese text image:

I tried English PDF and Images, and it work pretty well. Then I thought, why not I create a CLI tool, since it’s easier to do with LLMs. I used Claude Desktop App with Filesystem MCP Server, and it work pretty well. But I think it’s something I can improve to make it faster, like preparing proper prompt to generate a CLI that I want. I will create one and will share in a blog.

For commands, I created the following 3 commands…

  convert     Convert OCR JSON output to Markdown
  markdown    Process document and convert to markdown in one step
  process     Process a document with OCR

process will generate JSON file directly from Mistral. You can convert the JSON to markdown using convert command, or you can directly convert to markdown using markdown command directly. Mostly, I will personally use markdown command. It’s quite straight forward and easy. You can use –help flag to get helps. I added detail description inside the readme as well.