🤖 LinkedIn Easy Apply Automation Agent

Caution

This project is for educational and learning purposes only. Do not use this tool to spam job applications, violate LinkedIn's Terms of Service, or misrepresent your identity. The author takes no responsibility for any misuse.

An intelligent job application bot that reads your profile, opens LinkedIn, and fills Easy Apply forms automatically — powered by LangGraph + OpenAI GPT-4o-mini + Playwright.

---

System Architecture

graph LR
    You(["👤 You"])

    subgraph Portal["Profile Intelligence Portal"]
        UI["React UI<br/>:5173"]
        API["FastAPI<br/>:8000"]
        DB[("SQLite")]
        UI <--> API
        API <--> DB
    end

    subgraph Bot["Easy Apply Bot"]
        Agent["LangGraph Agent"]
        PW["Playwright Browser"]
        Agent --> PW
    end

    GPT["☁️ OpenAI GPT-4o-mini"]
    LI["🔗 LinkedIn"]

    You -- "chat to edit profile" --> UI
    You -- "⚡ Start Apply button" --> API
    API -- "reads profile" --> Agent
    API -- "chat LLM" --> GPT
    Agent -- "form analysis" --> GPT
    GPT -- "fill actions" --> Agent
    PW -- "automates" --> LI

Loading

---

What You Can Do in the Portal

Open http://localhost:5173 after starting both servers. The portal has two panels:

💬 Left — AI Chat (Profile Assistant)

Talk to the AI to build and update your profile:

What to say	What happens
"My name is Jane, email is jane@example.com"	Basic info saved to DB
"I know Python, FastAPI, Docker and AWS"	Skills list updated
"Expected CTC 10 LPA, notice period 30 days"	Job preferences saved
"Add a certification: AWS Solutions Architect"	Custom certifications key created
"Remove the certifications field"	Key deleted from profile
"What should I add to get more interviews?"	AI gives advice (no DB change)

The AI only writes to the database when you explicitly ask it to update something. You can ask questions freely — they won't modify your profile.

👤 Right — Profile Panel

• Live view of everything stored in your profile
• Custom fields (created via chat) appear under Additional Info with a × delete button
• Refresh button (↻) to reload from DB
• LinkedIn / GitHub / Portfolio quick links if set

---

Running Auto Apply

Option A — From the Portal UI (recommended)

1. Make sure both servers are running
2. Go to http://localhost:5173
3. Set up your profile via chat first
4. Click ⚡ Start Auto Apply in the top-right header
5. The button turns red and shows a live PID — the bot is running
6. Click ■ Stop Auto Apply to kill it at any time

The portal polls the bot status every 5 seconds. If the bot crashes, the button resets to idle automatically.

Option B — From the Terminal (headless / direct)

Important

When using Option B without the Profile Portal, the bot has no DB to read from. Before running, open user_profile.py and fill in EXAMPLE_PROFILE at the top with your own details:

EXAMPLE_PROFILE = {
    "name":   "Your Name",
    "email":  "you@example.com",
    "skills": ["Python", "FastAPI", ...],
    "expected_ctc": 900000,
    # ... all other fields
}

The bot reads this file at startup and uses it as context for every form field it fills.

# 1. Edit your profile first (only needed once):
#    open user_profile.py → edit EXAMPLE_PROFILE at the top

# 2. Login to LinkedIn (one time):
python data/setup_linkedin_browser.py

# 3. Run the bot:
python run_easy_apply.py

The bot opens Chromium, navigates to LinkedIn, finds Easy Apply jobs, and fills every form using your profile context.

---

Profile Context Priority

user_profile.py feeds the agent its context. It checks in this order:

1. SQLite DB  →  profile_portal/backend/profile_portal.db  (live, edited via portal)
2. EXAMPLE_PROFILE  →  user_profile.py  (fallback — edit with your details)

If running for the first time without the portal, open user_profile.py and edit EXAMPLE_PROFILE at the top with your own details — name, email, skills, CTC, etc. The bot will use this JSON as context for all form fields.

---

Step 3 — LangGraph applies, step by step

flowchart TD
    A(["▶ Open Easy Apply"]) --> B["extract_form_html<br/>Read modal step HTML"]
    B --> C["analyze_form<br/>GPT-4o-mini → List[FillAction]"]
    C --> D["execute_js_actions<br/>Playwright fills each field"]
    D --> E{"check_navigation"}
    E -- "Next →" --> B
    E -- "Review →" --> F["submit_application"]
    E -- "Submit →" --> F
    E -- "Stuck" --> G{"retry < 3 ?"}
    G -- "Yes" --> C
    G -- "No" --> H(["⏭ skip_job"])
    F --> I(["✅ Applied!"])

    style A fill:#7c3aed,color:#fff
    style F fill:#10b981,color:#fff
    style I fill:#10b981,color:#fff
    style H fill:#ef4444,color:#fff
    style C fill:#4f46e5,color:#fff

Loading

---

Chat Workflow

sequenceDiagram
    actor You
    participant UI as React UI
    participant API as FastAPI
    participant GPT as GPT-4o-mini
    participant DB as SQLite

    You->>UI: "Add TensorFlow to my skills"
    UI->>API: POST /api/chat
    API->>DB: load profile + history
    API->>GPT: profile JSON + chat history + message
    GPT-->>API: ChatReply { response, updates, delete_keys }
    API->>DB: update_profile(updates)
    API-->>UI: updated profile + reply text
    UI->>You: Shows reply, refreshes profile panel

Loading

---

Project Structure

job-apply-agent/
├── easy_apply_agent.py        # LangGraph agent (form fill logic)
├── run_easy_apply.py          # Entry point — Playwright loop
├── user_profile.py            # Reads SQLite profile → JSON for LLM
├── data/
│   └── setup_linkedin_browser.py   # One-time LinkedIn login
├── profile_portal/
│   ├── backend/               # FastAPI + SQLite
│   │   ├── main.py            # Routes (profile, chat, apply control)
│   │   ├── chat_service.py    # LLM chat + Pydantic structured output
│   │   └── db.py              # Profile & message CRUD
│   └── frontend/              # React + Vite
│       └── src/
│           ├── App.jsx
│           ├── components/ChatPanel.jsx
│           ├── components/ProfilePanel.jsx
│           └── api/client.js
├── requirements.txt
└── .env                       # OPENAI_API_KEY (gitignored)

---

Quick Start

# 1. Install
git clone https://github.com/akashcodejames/linkedin-easy-apply-agent
cd linkedin-easy-apply-agent
pip install -r requirements.txt
playwright install chromium

# 2. Set your API key in .env
echo "OPENAI_API_KEY=sk-..." > .env

# 3. Login to LinkedIn (one time)
python data/setup_linkedin_browser.py

# 4. Start the portal (two terminals)
cd profile_portal/backend && source venv/bin/activate && uvicorn main:app --reload --port 8000
cd profile_portal/frontend && npm install && npm run dev

# 5. Open http://localhost:5173 — update profile, then click ⚡ Start Auto Apply

---

Tech Stack

	Tech
Browser	Playwright (Chromium, persistent session)
Agent	LangGraph state machine
LLM	OpenAI GPT-4o-mini
Structured output	Pydantic v2 + with_structured_output()
Profile store	SQLite → json.dumps into LLM prompt
Backend	FastAPI + uvicorn
Frontend	React + Vite
Chat memory	ConversationSummaryBuffer