Star on GitHub

Job hunting is broken.

Applying to 500+ jobs manually is soul-crushing. Existing tools just spam untailored resumes blindly. We built a system that actually parses the job, rewrites your resume, and bypasses complex forms using pure visual AI.

How It Actually Works

01

Drop Your Resume

Upload your master resume once. AutoClaw parses it perfectly, maps your skills, and builds a centralized data profile requiring zero manual data entry.

02

Agent Scours Web

The Scanner browses job boards autonomously. It extracts listings, scores them against your profile out of 100, and filters out the noise.

03

Tailor & Apply

AutoClaw tailors your resume for that specific job, drafts a cover letter, and visually clicks through ATS forms using Gemini 2.0 without breaking.

04

Interview Prep

It detects interview invites in your email, syncs them to your calendar, and generates a detailed research brief 48 hours before you speak.

Pipeline Tracker

Watch the magic happen live. A beautiful Kanban board that tracks every application's status from 'Drafted' to 'Interview Scheduled'.

Frontend Engineer • Foundit
Score: 92/100
React Dev • Internshala
Interview Scheduled

Q&A Memory Vault

Tired of answering the same questions? AutoClaw remembers your best responses to screening questions and dynamically adapts them for future applications.

MEMORY ACCESSED
"Recalling previous successful answer about Next.js experience... Tailoring specifically to match current company core values."

How AutoClaw Actually Works

Beyond simple CSS selectors. Engineered for adversarial DOM manipulation and complex ATS platforms.

Vision-Language-Action Fallback

When standard CSS selectors fail on obfuscated DOMs, AutoClaw dynamically injects set-of-mark bounding boxes on the page instead of giving up.


A viewport screenshot is captured and piped to the Gemini 2.0 Flash Vision API. It infers semantic meaning and executes pixel-accurate hardware-level clicks to achieve a 100% bypass rate on adversarial platforms.

Sub-1s
Vision API Latency
100%
DOM Bypass Rate
autoclaw-core.py
# [AutoClaw Phase 4: Execution]
INFO: Planner Agent starting task: Apply to Frontend Dev
PERF: Loading tailored resume from PostgreSQL vault.
# Initializing Applicant Agent (Vision-First)
INFO: Injecting Set-of-Mark JS payload on job form...
SUCCESS: 42 Bounding boxes drawn.
INFO: Transmitting viewport dump to Gemini 2.0 Flash Vision...
> Vision Inference Complete
{
"target_id": "box-17",
"confidence": 0.992,
"coordinates": {x: 842, y: 1105}
}
INFO: Executing hardware-level click at (842, 1105)
SUCCESS: Transitioning to Phase 1: Authentication

Planner Agent

The master brain. Scores jobs, delegates tasks, and prevents loops.

Scanner Agent

The researcher. Browses job boards and extracts requirements via Vision LLM.

Applicant Agent

The executor. Instantly tailors resumes and fires the Validator loop dynamically.

Outsmart The System

Engineered With

Next.js 14
Tailwind CSS
FastAPI
Playwright
PostgreSQL
Redis
Gemini 2.0 Flash
Docker
Next.js 14
Tailwind CSS
FastAPI
Playwright
PostgreSQL
Redis
Gemini 2.0 Flash
Docker
Next.js 14
Tailwind CSS
FastAPI
Playwright
PostgreSQL
Redis
Gemini 2.0 Flash
Docker
Next.js 14
Tailwind CSS
FastAPI
Playwright
PostgreSQL
Redis
Gemini 2.0 Flash
Docker