Set up
Install it with pipx, run jobwatch and follow the four onboarding steps. Your page opens on Today.
Your job hunt shouldn't be a full-time job.
jobwatch watches company job boards, ranks new roles, tracks applications and fills application forms, and runs on your own computer.
pipx install jobwatchThen run jobwatch. Or pip install jobwatch. MIT licensed.
The whole app in a minute and a half, made from made-up data.
Six steps, each with a short video made from made-up data.
Install it with pipx, run jobwatch and follow the four onboarding steps. Your page opens on Today.
jobwatch watches the company boards you pick. Today ranks the new roles, optionally with a fit score against your resume.
A short queue of jobs to apply to, with a resume tailored per job. The Chrome extension fills application forms from your details and earlier answers. You press Submit.
Every stage through to an offer, with the next step and a follow-up day. Status emails update it, and applications that go quiet get flagged.
A prep sheet per interview, built from the resume and answers you sent, with answers drafted from your own stories, a mock interview and a drill video.
Compare pay, career value and company health side by side, and see who you know at each company.
Greenhouse, Lever, Ashby, Workable, Workday and more, plus HN's Who is hiring and VC portfolio boards. Fresh roles, real pay ranges, robots.txt respected.
Only new roles that match you, optionally fit-scored against your resume: must-haves met, gaps, a score out of 100.
A short queue of jobs to apply to, with a resume tailored per job. The Chrome extension fills the form from your details and earlier answers.
Every stage, the next step, follow-ups for applications that go quiet, a calendar feed, and offers side by side.
A prep sheet per interview, answers drafted from your own stories, a mock interview and a drill video.
An MCP server and an agent skill, so Claude and other assistants can run your search with you.
Where the AI runs: the Claude API, Antigravity or this Mac, the fallback chain, and what leaves your computer.
Sized by how much memory your computer has. jobwatch's page suggests one for yours.
| Memory | Mac with Apple Silicon (MLX, runs on the GPU via Metal) | Windows, Linux or Intel Mac (Ollama) | Also good: Gemma 4 (MLX first, then Ollama) |
|---|---|---|---|
| 8 GB | Qwen3.5-4B, 4-bit Qwen3.5-4B-MLX-4bit ↗suggested, sized by memory |
qwen3.5:4b-q4_K_M ↗ | gemma-4-e2b-it-4bit ↗ gemma4:e2b-it-q4_K_M ↗ |
| 16 GB tested on a 16 GB Mac |
Qwen3.5-9B, 4-bit Qwen3.5-9B-MLX-4bit ↗ |
qwen3.5:9b-q4_K_M ↗same model; the test ran the Apple Silicon build | gemma-4-12B-it-4bit ↗ gemma4:12b-it-q4_K_M ↗tested: most accurate (29/29 vs 28/29), about 50% slower, +1.7 GB |
| 24-32 GB | Qwen3.8-27B, 4-bit Qwen3.8-27B-4bit ↗suggested, sized by memory |
qwen3.8:27b-q4_K_M ↗ | gemma-4-26b-a4b-it-4bit ↗ gemma4:26b-a4b-it-q4_K_M ↗ |
| 48 GB and up | Qwen3.6-35B-A3B, 4-bit Qwen3.6-35B-A3B-4bit ↗ 8-bit with 64 GB or more: Qwen3.6-35B-A3B-8bit ↗suggested, sized by memory |
qwen3.6:35b-a3b-q4_K_M ↗ 8-bit with 64 GB or more: qwen3.6:35b-a3b-q8_0 ↗ |
gemma-4-31b-it-4bit ↗ gemma4:31b-it-q4_K_M ↗ |
On computers with 48 GB or more, 35B-A3B-style models (about 3B active parameters) are the fastest choice.
On an Apple Silicon Mac, use MLX: it's built for Apple's chips and runs faster than Ollama there. Ollama is for computers MLX doesn't run on.
Only the 16 GB row was tested: on a 16 GB Apple Silicon Mac it agreed with 28 of 29 expected verdicts, at 27 seconds a call and 7.5 GB of memory. The other rows are suggested, sized by memory. The Gemma 4 column is an alternative for each row; on Apple Silicon it is served the same way as the Qwen beside it.
jobwatch sends nothing home, so stars are the only count of who uses it.
Star on GitHub