JobRadar: A local LLM agent for job hunting that doesn't make decisions for you
Sh0ny
2 августа 20261 min read
In short
This open-source utility collects job listings from eight sources and ranks them using a local model—without the cloud or subscriptions. But the author’s main point is that AI serves only as an advisor here, and the final decision rests with the user.
JobRadar is an open-source CLI agent that simultaneously polls eight free job sources: Remotive, Arbeitnow, RemoteOK, Jobicy, Himalayas, Greenhouse, Ashby, and, optionally, LinkedIn. Instead of opening hundreds of tabs, it filters out the noise and provides a ranked list with brief notes explaining why a job is a good fit or not.
The main architectural focus is a local LLM. JobRadar automatically detects Ollama or llama.cpp and evaluates each job opening on a scale of 0–100 across four dimensions: skill match, experience relevance, salary, and remote work. No cloud APIs or subscriptions: everything is processed on the user’s device. For those who don’t want to send their resume to a third-party cloud, this isn’t just a nice-to-have—it’s a basic requirement.
But the most interesting aspect is its positioning. The README states explicitly: AI is a guide, not a decision-maker. The model doesn’t understand your feelings about a company’s culture or the fact that you’re willing to work on a smaller team for less money. These assessments are a starting point, not a verdict.
This is an honest trade-off that’s rarely acknowledged in products labeled “AI-powered.” A local model can’t replace the context you keep in your head, but it can take the drudgery out of the initial screening—especially if there are hundreds of job openings and only eight sources.
On the practical side: there’s a cache so you don’t have to review the same job postings every time you run the tool, and a web dashboard at localhost:3000 with a Kanban pipeline, filters, and a configuration editor. Installation is via npm, a curl script, or PowerShell.