AI browser automation for job hunting used to mean paying for a SaaS or fragile extension. Now your own agent — Claude Code or Codex — can find roles, fill approved fields, and return a reviewable application list with ego (lite) for free.
We’ll use LinkedIn as the example of AI browser automation for job hunting. The same steps work on any board your agent can open — you just change which roles you ask for.
A Chromium browser you use daily. One click imports your Chrome logins, so the agent applies as you.
Which roles should it apply to?
/ego-browser Find the 10 newest remote React roles on LinkedIn and prepare applications using my profile
Type /ego-browser and just say which roles — in any language your agent understands.
Get a log of what it applied to — or have it draft each one for you to review before it submits.
Point your agent at a board you are authorized to use and say which roles to search or apply to. The workflow is similar, but each board's forms, login checks, and rules still apply.
“Apply to the 10 newest remote React roles, using my profile.”
“Apply to marketing roles in NYC posted this week with my resume.”
“Fill and submit this Greenhouse application from my profile.”
“Complete this Lever application using my saved answers.”
“Walk the Workday flow and fill every step from my resume.”
“Apply to senior PM roles I haven't applied to yet.”
“Submit the Ashby application with a cover letter tailored to the role.”
“If you can apply there, name it and let it run.”
The useful part of job-application automation is removing repeated typing while keeping decisions, account access, and truthfulness with you. Use these guardrails when you scout roles, fill an ATS form, or prepare a submission.
Before Submit is the control point: compare the role and company, confirm the selected resume, read every answer, and check work authorization and salary fields. Require the agent to pause on missing or ambiguous information so a fast run never turns an assumption into a claim.
Import an authorized browser profile and keep account state isolated in its own Space. The agent can reuse a signed-in session, but you remain responsible for MFA, CAPTCHA, identity checks, and any re-authentication that the job board requests.
Greenhouse, Lever, Workday, and other ATS flows can add conditional fields or change labels without notice. Ask the agent to describe each unfamiliar field, verify required attachments, and update your answer bank after a reviewed run instead of relying on brittle selectors or an old template.
A scouting pass can collect titles, employers, locations, URLs, and fit notes into Markdown or CSV before any application is opened. Paste approved rows into a Google Sheet in your own browser only after checking the sheet, sharing settings, and duplicate history.
Define seniority, location, employment type, visa constraints, salary range, required skills, and exclusions up front. Log why each posting was included or skipped so you can tune relevance without encouraging a batch of low-quality submissions.
The agent may reorder and rephrase facts you supplied, and it can produce a plain-text version with conventional headings for ATS parsing. Review every edit and reject invented credentials, dates, employers, metrics, or responsibilities; tailored wording must remain evidence you can defend in an interview.
Some teams welcome AI-assisted drafting, while others ask applicants not to use it. Follow the employer's instructions, keep each application specific to the role, and disclose assistance when requested; a human-reviewed submission is more defensible than an identical mass blast.
Rate limits, CAPTCHAs, identity checks, timed assessments, and unverifiable job listings are boundaries, not obstacles to route around. Stop the Space, verify the employer and destination URL, and leave the item as needs review while you follow the platform's terms and applicable privacy rules.
Hand it your resume and criteria, then choose whether it should scout openings, prepare applications, or pause for your review. A completion notice points you to the source-linked results; the visible Space remains available while it works.
| Company | Job title | Location / work type | Direct application link |
|---|---|---|---|
| AutoComplete | AI Product Engineer | San Francisco, CA · Full-time · On-site | Apply on LinkedIn |
| Comulate | AI Engineer | San Francisco, CA · Full-time · On-site | Apply |
| UCSF Health | AI Engineer (AI Accelerator Program) | San Francisco, CA · Full-time · On-site | Apply |
| Metriport | AI Engineer | San Francisco, CA · Full-time · On-site | Apply |
ego (lite)Task complete — 10 open roles found, links readyThe follow-up costs one more line: “Now apply to all 10 with my profile.”
Auto-applying usually means signing up for a service per board, installing an extension that breaks, or wiring up your own script — and paying to keep it alive. You skip the entire stack.
npm install puppeteer-core and a Node project to maintaininput, data-testid) that break when the board’s UI changesBuilt to apply on your behalf, with your own profile, on accounts you already have access to.
Apply on your behalf, with your access.
The red lines, stated plainly.
No. You bring an AI agent you already use — Claude Code, Codex, or Cursor — and ego (lite) is the browser it drives. The agent applies on your own logged-in session, so there's no job-board API to apply for, no API keys, and no third-party auto-apply service in the loop.
An AI browser agent can search the job boards you already use, filter roles against your criteria, collect direct application links, and fill repetitive fields from an answer bank. In ego (lite), the work runs in a visible browser on your Mac, so you can inspect the posting and review the completed application before anything is submitted.
It can prepare and submit a bounded set of applications when you explicitly ask it to, but the safer workflow pauses before each submission for your review. Set a small daily cap, stop on unfamiliar questions or verification screens, and never treat the page as a promise of unlimited or unattended applications.
The agent can return a source-linked job list or application log with the role, company, location, job URL, status, and check time. That record lets you review what was found, submitted, skipped, or blocked instead of losing track of applications in a browser tab.
ego (lite) is a free download. Instead of a monthly subscription or per-application credits, you only spend your own agent's tokens — and a single application usually costs just a few, since the agent works from a compressed snapshot of the page rather than raw HTML. There's no third-party service to pay or depend on.
LinkedIn, Indeed, Greenhouse, Lever, Workday, Glassdoor, Ashby, and other boards whose visible application flow your agent can open. The workflow is similar across them, but each board's forms, login checks, access, and rules still determine what can be completed.
No. It applies with your real profile and your own answers, and it tailors the wording of your resume and cover letter to each role — it doesn't fabricate qualifications, degrees, or work history. If you'd rather check first, tell it to draft each application for your review before it submits.
ego (lite) runs on your real, logged-in session and behaves like you applying yourself. It doesn't defeat CAPTCHAs, evade identity checks, or get through timed assessments, and it doesn't mass-blast recruiters. We can't promise any platform's rules won't change, so stay within each platform's terms and use it the way you'd apply by hand.
Your browsing data stays on your own computer — ego (lite) doesn't upload your history, cookies, sessions, or resume. It's free on macOS today, with Windows and Linux on the roadmap. If you connect an external agent or model provider, that provider has its own data policy.
Give the agent a written answer bank, require it to stop on anything it cannot answer, and review the complete form before every submission. A good workflow checks the role, company, location, resume version, screening answers, and work-authorization response as one checklist; it never treats a plausible-sounding generated answer as permission to submit.
Yes. ego (lite) drives a browser context you already own and are signed into, so the agent can reuse that session without receiving your password. Import only the profile you authorize, keep separate Spaces for separate accounts, and take over whenever a login, MFA, CAPTCHA, or identity check appears.
It can work through visible fields and steps on ATS forms that your session can open, including forms that reveal questions dynamically. ATS layouts change frequently, so treat every run as supervised maintenance: stop when a field is unclear, verify attachments and required fields, and update your answer bank when a company adds a new question.
The agent can scout boards, deduplicate listings, and return a CSV or Markdown table with the title, company, URL, location, and status. You can then ask it to open an approved Google Sheet in your own signed-in browser and paste the rows, but review permissions and the columns before writing; the page does not promise a native Sheets integration or unattended syncing.
Put hard filters and explicit exclusions in the prompt, then require a role-by-role fit check before any form is opened. Titles, seniority, location, employment type, visa requirements, salary range, and duplicate history should be recorded in the log so you can see why a posting was included or skipped instead of optimizing for application count.
It can reorder and rephrase facts you supplied to match a role's terminology, while preserving your real dates, employers, skills, and outcomes. Tell it to flag missing evidence, keep a plain-text ATS-safe copy, and show every proposed change for approval; it must not add a degree, certification, metric, or job responsibility you did not provide.
Recruiter reactions vary by company and role, so automation is not a substitute for relevance or a thoughtful application. Keep the submission truthful and specific, avoid identical mass-blasts, follow each employer's instructions, and disclose AI assistance when a form or recruiter asks; human review is what keeps the application attributable to you.
Treat rate limits, suspicious listings, blocked pages, and verification screens as stop signals. Do not bypass controls or submit to a posting you cannot verify; cross-check the employer and destination URL, respect the job board's terms and privacy rules, keep a conservative cap, and leave anything uncertain in the needs-review state.