I built dot-mcp: it gives ChatGPT a computer (shell, files, memory, browser)

I am the creator of dot-mcp, an open-source MCP server written in Python. To get started, simply install it using pip install dot-mcp. Once installed, dot-mcp runs on your own machine, providing ChatGPT, Claude, or Gemini with a range of tools. These tools include a shell, file access, persistent memory, and a browser, totaling 46 tools in total. You can see the server in action in the 44-second demo, which is a real screen recording and not a mockup (it plays at the top of the README): https://github.com/Ashveil1/dot-mcp.

In the demo, ChatGPT with a free plan and a custom connector opens Hacker News using the browser tool, reads the top five stories, writes a Python script that saves the stories in a ranked table format, and runs the script. The file path is then saved to memory under a specific tag. When a new chat begins, the assistant can recall that note from memory. The memory is stored on disk, so it survives a server restart.

MCP allows a chat model to call tools, and dot-mcp serves as a single local server that bundles the basic ones: running commands, reading and writing files, remembering things, and driving a browser. This enables a chat assistant to perform real work on a computer instead of merely discussing it.

However, it's important to note that the server grants full shell access to anyone who holds a credential. Therefore, you should only expose it while you are using it and stop it with dot-mcp stop once you're done. For more information, you can visit the project's repository at https://github.com/Ashveil1/dot-mcp or install it directly from PyPI at https://pypi.org/project/dot-mcp/.

Please keep in mind that the project was developed by me and the demo video was created with the assistance of an AI assistant. The article about dot-mcp was also posted by the AI on my behalf. I welcome any feedback and bug reports to help improve the project.