Fine Tune Job Agent

Stop reading every job description yourself.

Give your AI coding agent your background, your standards, and one clear job-search goal. FTJA finds the LinkedIn roles worth your attention — and shows you why.

See how it works
The value

A job search shaped around your judgment.

01
Your background
Add your resume. The file stays on your computer, your own AI agent reads it, and you confirm the profile it drafts.
Resume, portfolio, and conversation flow into a profile
02
Agent judgment
Evaluate public LinkedIn jobs against criteria you can see and edit.
Scraping flows into code and then through stage 1 and stage 2 agents
03
Better next searches
Your apply/skip decisions and reasons are used to make the next search better.
Profile and agent connected in a learning loop
How it works

Fine-tune your job agent.

Your agent guides you in chat. You decide in the web view: add your background, then review the profile, search settings, and judgment rules behind every run. Your decisions keep shaping the next one.

DEMO/ftja-setup
Add your background.
Add the resume FTJA should read to understand your background.
Your source filesFTJA will use it to prepare a concise local profile.
+ Add resumePortfolio · after beta
A real run

Twenty seconds of an actual search.

Sped up. One run over four job titles.

1,991postings scraped
396past the code-based filter
45passed, each with quoted evidence
Where this is going

Search is the first step.

Three things FTJA is heading toward, in order.

01 · Next

Your agent, more places to look.

FTJA runs in Claude Code and reads LinkedIn today. Next it comes to more agents, including Codex and Hermes Agent, and to more sources: Indeed, and the career pages companies run on Greenhouse, Lever and Ashby.

Agents on the left (Claude Code now; Codex and Hermes Agent next) and sources on the right (LinkedIn now; Indeed and company career pages next), all connected to FTJA
02 · Then

Applications, drafted for you.

When a role is worth applying to, your agent prepares the application: your resume attached, the motivation written, each question answered. It draws on your profile, on what it has learned from your conversations, and on the job posting itself.

You review it and confirm. Nothing is sent without you.

Your profile, what you have told your agent, and the job posting flow into an application form that ends with a Review and confirm button
03 · Vision

No one applies. No one posts a job.

Companies get an agent too. It holds what a job post cannot: what the team needs right now, and the preferences nobody publishes. Your agent holds your judgment, taught run after run.

The two agents talk, and each side gets a short, curated list. The right people still meet.

A direction, not a schedule.

Your agent, holding your private judgment, talks with a company's agent, holding what the team needs now and its unpublished preferences; both sides receive a short list
Not a black box

See how your agent reaches a decision.

Public sources

FTJA scrapes public LinkedIn listings without a login. LinkedIn controls coverage, so FTJA states that limit plainly.

Each judgment step

See what was scraped, what the filters removed, which evidence matched, and why a role made it through.

Local files. Open code.

Your profile, criteria, run history, and results stay in local files. FTJA is open source, so you can inspect and change the pipeline itself.