boss-agent-cli

Why we included this project

Anyone building AI agents that help with job hunting on China's BOSS Zhipin platform will find this fits neatly into that workflow. Instead of trying to automate the risky actions that job boards block, it exposes read-first capabilities: keyword search with multi-dimensional and welfare filtering, cached job detail lookup, and a local shortlist with tags and funnel stats, all returning structured JSON envelopes that Claude or Cursor can call through MCP. The schema-driven design lets the agent read its own command definitions, and an assisted operating mode blocks sensitive actions like greeting or applying by default, so you stay on the compliant side of the platform. That division of labor, where the tool handles discovery and organization while you manage the human interactions in the browser, is what makes it usable in real daily workflows rather than just a demo. Because the output is machine-readable, you can also feed results into a larger pipeline for ranking or comparison.