Terminal-first job-search assistant. Filter, save queries, and (optionally) rank with an LLM.

job-kick is a CLI for skimming 25 job postings without 25 browser tabs. Search, filter, and save queries from the terminal — bring an LLM key for ranking and per-job analysis.

$ jobq search --source linkedin -k python -l Germany --remote-only
                       LinkedIn — 12 jobs
┏━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ # ┃ Job ID   ┃ Title            ┃ Company    ┃ Location ┃ Posted     ┃
┡━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━┩
│ 1 │ 12345678 │ Senior Python    │ Acme       │ Remote   │ 2026-04-29 │
│ 2 │ 12345679 │ Backend Engineer │ Globex     │ Berlin   │ 2026-04-28 │
│ 3 │ 12345680 │ ML Engineer      │ Initech    │ Remote   │ 2026-04-27 │
└───┴──────────┴──────────────────┴────────────┴──────────┴────────────┘

Job search can be exhausting — too many sources, too many irrelevant postings. job-kick is a terminal-first job-search assistant: search a source, filter by keyword, location, recency, remote, and job type, and save queries for re-use. No login, no algorithm picking what you see.

Bring an LLM key (Anthropic or OpenAI) if you want the extras: natural-language queries, profile-based ranking, and per-job analysis. None of these are required to use the core.

job-kick isn't on PyPI yet. For now, install directly from the GitHub repo (Python 3.11+):

pip install git+https://github.com/kasun/job-kick.git
# or with uv:
uv tool install git+https://github.com/kasun/job-kick.git

jobq configure

Or clone and install for local hacking:

git clone https://github.com/kasun/job-kick.git
cd job-kick
uv sync           # creates .venv and installs deps
uv run jobq configure

jobq configure walks through three steps: default source, LLM provider (Anthropic or OpenAI — bring your own API key), and search profile. The LLM features (--prompt, --match, match, summarize) need a key; everything else works without one.

# search by keyword and filters
jobq search -k python -l Germany --remote-only --source linkedin

# save a query for re-use
jobq search -k python -l Germany --remote-only --save-template eu-python
jobq search --template eu-python

# bookmark interesting jobs (job ids come from the table)
jobq bookmarks add 12345678 --source linkedin

# natural-language search → structured filters
jobq search -p "remote python jobs in EU posted this week"

# rank every result against your profile
jobq search --template eu-python --match

# deep-read one job — summary, fit, red flags, interview questions
jobq match 12345678

Run jobq <command> --help for full options. -v / --verbose (before the subcommand) enables debug logging to stderr.

Run jobq profile edit to open a markdown file in $EDITOR. Write what you're looking for in your own words — when you opt in to ranking with an LLM key, this file is the rubric:

# What I'm looking for
- Remote senior Python roles, EU timezones
- Comp band €80-110k
- Small teams, no on-call

# About me
- 8 years backend, currently in Berlin
- Strong in Python/Postgres, light on AWS/k8s

Treat it as a journal, not a CV. The richer the intent, the better the scores.

• Local-first. No telemetry. The only network calls are to the source you configure (e.g. LinkedIn) and the LLM provider you configure.
• Everything lives under ~/.config/jobq/:
  • config.toml — settings
  • credentials.toml — API keys (file mode 0600)
  • profile.md — search profile
  • data.json — bookmarks and templates
• Delete the directory to start over.

• Sources are pluggable. LinkedIn (public job pages, no auth) is the only one in 0.1. Adding a source means implementing the JobSource protocol in src/job_kick/sources/base.py.
• LLM providers are pluggable via LiteLLM. Anthropic and OpenAI are first-class. Switching to a local Ollama, Bedrock, or anything else LiteLLM supports is a config change.
• Storage is a single human-readable JSON file (TinyDB). Inspect it, edit it, version-control it if you want.

jobq is v0.1 — early. What works today:

• LinkedIn public-jobs search and job fetch (no login required).
• LLM-backed prompt parsing, single-job match analysis, batch match scoring, summarization.
• Local bookmarks, saved search templates, search profile.

What's coming:

• Authenticated LinkedIn (richer filters, fewer rate limits).
• Mark seen jobs.
• Scheduling, notifications.
• Additional sources, open an issue if you would like a specific source.

Issues and PRs welcome — small, scoped changes land faster.

MIT — see LICENSE.