OpenComputerUse: driving computer use with Jev
The open-source OpenComputerUse project supports background windows, lock screens and MCP, and its Recipes feature uses Cloudflare Clev or Jev models to walk the accessibility tree and execute an ordered list of actions sent by an agent.
Introducing OpenComputerUse - best in class computer use for any agent Traditionally, only ChatGPT Desktop was blessed with great computer use. Introducing OpenComputerUse! It supports the following: - Background usage of windows - Beautiful halo and cursor view like ChatGPT Desktop - Recipes - Use @Cloudflare's Clev or @typesafeai's Jev models to work with your agent. Your agent can send a list of things to do in order, and presuming none of them need a LLM, these models can walk the a11y tree and get the job done! - Lock screen working - Usable with all agents with the MCP - There is a built in HTTP API you can optionally turn on and within the app generate a skill that will with the filled in HTTP route to your system/what your system is to it make a skill to connect to it and do actions! In the video attached, you can see it doing computer use and also handling stumbling blocks with Firefox in a new profile and cookie consent! https://github.com/IAmJSD/OpenComputerUse