AIHawk is a Python terminal application that reads a resume, searches a published index of more than ten million job postings for titles that fit, scores the best matches against that resume, and then fills in application forms in a browser the user can watch — built for individual job seekers rather than recruiting teams.

What it is

AIHawk, published as Jobs_Applier_AI_Agent_AIHawk under the MIT license, is a Python application that runs in the terminal and needs Python 3.11 or newer. It reads a resume, writes a search plan from it, takes the applicant's contact details from it, and uses that plan to search a job database published as a release. It sits in the Python AI-agent and browser-automation ecosystem: form filling is carried out through invisible-playwright-mcp, and the project carries the ai-agent, browser-automation, mcp, job-search and terminal-app topics.

The problem it solves is the manual loop around job hunting — trawling boards for titles that fit, judging each posting against a resume, then retyping the same details into form after form on different applicant tracking systems. AIHawk replaces that loop with a local workflow: a one-time index download, a title-based shortlist, a 1-to-5 score per job with its gaps, and then form filling in a visible browser that stops short of submission until the user confirms.

Key capabilities

• Reads a resume supplied as a PDF, a .docx or a text file and derives both a search plan and contact details from it; a scanned PDF has no text to read, so it must be exported with text or saved as .docx.
• Searches a job database published as a release — 10,223,753 postings in the manifest of 2026-11-04, with postings older than 60 days left out by default — downloaded once per database at 672 MB into ~/.aihawk/data/index.
• Shortlists the best 100 postings by title, has the Jev decision model answer six questions about each, keeps the best 25, and scores each from 1 to 5 against the resume with the gaps and reasons shown on Enter.
• Drives a terminal table with single-key actions: space to select, a to send after one yes, s to list the jobs sent and what each page said, o to open a posting in the browser, m to score more jobs, r to give a new resume, l to open the browser for sign-in.
• Keeps a human approval gate in the apply path: the model fills the form and names the submit button but is told not to send, and clicks that would send are refused until the user presses a and confirms.
• Covers twelve applicant tracking systems through plugins — Ashby, BambooHR, Breezy, Greenhouse, JazzHR, Lever, Paylocity, Personio, Recruitee, SmartRecruiters, Teamtailor and Workable — plus read-only Gmail and Outlook plugins that take a code or a link from a mailbox.
• Adds a site by writing a Markdown file in src/aihawk/plugins/ with no code required, and keeps private plugins in ~/.aihawk/plugins/; each send is recorded in applied.csv and data/proofs/.

Who uses it and how

• Individual job seekers running aihawk resume.pdf on their own machine and working through the resulting table, approving each send one at a time rather than letting the agent submit unattended.
• Applicants reaching across many applicant tracking systems at once, where the plugins cover the common forms and postings on sites without a plugin are still listed for applying by hand.
• Users who want to watch and intervene: the browser is visible throughout, sign-in is done by hand through l, and whatever a form asks that the files do not answer is put to the user and remembered.
• Contributors extending site coverage, who add a Markdown description of one site's form either to the repository or to their local plugin directory.
• People working from a local index: after the one-time 672 MB download, a search reads the copy on the computer, while the descriptions of the jobs being judged are read from the release over the internet.

Getting started

Install with pipx install git+https://github.com/feder-cr/Jobs_Applier_AI_Agent_AIHawk, or with uv tool install followed by uv tool update-shell and a new terminal, or into a virtual environment. Then run aihawk resume.pdf; the first start asks for an OrcaRouter key, kept in ~/.aihawk/key, with ORCAROUTER_API_KEY overriding that file.

How it compares

No list of paid products is supplied for this project, and the facts name no sibling tool of the same kind, so within this registry it stands alone as a terminal-driven, MIT-licensed job application agent. The one closely related component named in the documentation is invisible-playwright-mcp, which is the browser-control layer AIHawk uses rather than an alternative to it.

When to use it — and when not to

The tool depends on an OrcaRouter key, so it is not a self-contained offline system, and the first start pulls a 672 MB index while the daily refresh workflow, jobs-database.yml, is on main and has not run yet, so index freshness is unproven. The Gmail and Outlook plugins have not been tried on a real mailbox, and the documentation warns that as a last resort the model can run a script that could send a form without asking, which is why SECURITY.md matters. Anyone wanting unattended bulk applications, or unwilling to hold an OrcaRouter account, should not pick it — the design deliberately requires one human yes per send.

pipx install git+https://github.com/feder-cr/Jobs_Applier_AI_Agent_AIHawk

python -m venv aihawk-env
aihawk-env/bin/pip install git+https://github.com/feder-cr/Jobs_Applier_AI_Agent_AIHawk
aihawk-env/bin/aihawk resume.pdf

 AIHawk  Ada Rossi · Python Engineer, Backend Developer · Milano, Italy      [scoring 12/40]
 ┌───┬───────┬───────┬──────────┬─────────────────────────┬──────────┬────────┐
 │   │ Score │ Apply │ Company  │ Role                    │ Location │ Status │
 │ ● │ 4.8   │ auto  │ Acme     │ Senior Python Engineer  │ Milan    │        │
 │   │ 4.5   │ auto  │ Beta     │ Backend Developer       │ Remote   │ sent   │
 │   │ 4.1   │ link  │ Gamma    │ Python Developer        │ Rome     │        │
 └───┴───────┴───────┴──────────┴─────────────────────────┴──────────┴────────┘