blacksuan19/h1b-mcp

h1b-mcp is a server that gives AI coding agents, such as Claude Code or Codex, the ability to check whether a company sponsors H-1B work visas in the United States. It plugs in using the Model Context Protocol, a standard way for AI tools to call external services, and works by opening the H1BGrader website in a real browser, searching for a company, job title, or location, and pulling out the information shown there. This project is an independent community effort and is not affiliated with or endorsed by H1BGrader itself. Once connected, an agent can ask for a plain sponsorship signal like strong, mixed, weak, or no sponsor for a given company, along with supporting detail such as yearly visa application approval and denial counts, salary ranges, and an overall grade that H1BGrader assigns to that employer. The same kind of lookup also works for job titles and for specific cities, showing which employers most commonly sponsor that role or hire in that location. Results are cached locally so that asking about the same company again does not require reopening the browser and re-scraping the page. If the website blocks the automated browser or a lookup otherwise fails, the tool raises a clear error back to the agent rather than guessing or inventing an answer, so the agent knows the information could not be retrieved. Setting it up requires installing the uv Python tool so that its uvx command is available, then adding a short configuration snippet to your AI client of choice, such as Codex, Claude Desktop, Claude Code, or VS Code, pointing it at the published h1b-mcp package. Once configured and the client is restarted, you can simply ask your AI assistant something like whether a particular company sponsors H-1B visas, and it will call this tool automatically to answer. The project is released under the MIT license.