GitHub - Vellorpavan/AUTONOMOUS-LEETCODE-AGENT: 🤖 An autonomous AI agent that logs into your LeetCode account and independently solves problems using a custom MCP server.

██╗     ███████╗███████╗████████╗ ██████╗ ██████╗ ██████╗ ███████╗
██║     ██╔════╝██╔════╝╚══██╔══╝██╔════╝██╔═══██╗██╔══██╗██╔════╝
██║     █████╗  █████╗     ██║   ██║     ██║   ██║██║  ██║█████╗  
██║     ██╔══╝  ██╔══╝     ██║   ██║     ██║   ██║██║  ██║██╔══╝  
███████╗███████╗███████╗   ██║   ╚██████╗╚██████╔╝██████╔╝███████╗
╚══════╝╚══════╝╚══════╝   ╚═╝    ╚═════╝ ╚═════╝ ╚═════╝╚══════╝
     A U T O N O M O U S   A G E N T   ×   M C P   ×   A I

🤖 One command. Every LeetCode problem. Zero human input.

Built on Google Cloud · Powered by Claude Opus 4 · Custom MCP Protocol

📺 [Click to Watch] Autonomous LeetCode Agent in Action!

📌 Table of Contents

Click to expand / collapse

🧠 What Is This?
✨ Features
🏗️ Architecture
📁 Project Structure
⚡ Quickstart
💻 Usage
🛠️ MCP Server — Tool Reference
🤖 Agent System Prompt
📊 Session Logs
☁️ Google Cloud Deployment
🐛 Troubleshooting
🗺️ Roadmap
🤝 Contributing

🧠 What Is This?

LeetCode Autonomous Agent is a fully automated AI system that logs into your LeetCode account and solves every single problem — on its own, without any human input after you press start.

It combines three things:

A custom MCP Server (Model Context Protocol) built from scratch — wraps LeetCode's private GraphQL API as clean, callable tools
Claude Opus 4 as the AI brain — reads problems, reasons about algorithms, writes optimal Python 3 code
An autonomous agent loop — manages the problem queue, submits solutions, reads verdicts, fixes failures, and moves on — forever

You start it once. It runs until your LeetCode profile is solved.

You
 │
 └─▶ ./run.sh solve-all
         │
         ▼
    Agent starts
         │
    ┌────▼────┐
    │  Queue  │ ← 3000+ problems
    └────┬────┘
         │
    ┌────▼────────┐
    │  Read prob  │
    │  Write code │
    │  Submit     │
    │  Check      │
    │  Fix/retry  │
    └────┬────────┘
         │
    ┌────▼────┐
    │  Next   │──▶ loop forever
    └─────────┘

✨ Features

Feature	Description
🤖 Fully Autonomous	Zero human input after start. The agent manages the entire loop
🧠 Elite Solver Prompt	10-step competitive programming workflow built into the system prompt
📋 Smart Queue	Fetches all problems, sorts Easy → Medium → Hard, skips already-solved
🔄 Verdict-aware Retry	Different fix strategy for WA, TLE, MLE, RE, CE — not the same blind retry
🛡️ Rate Limit Guard	Auto-waits on 429, retries once, skips and continues if persistent
📊 Session Logging	Every problem outcome logged to JSON with timestamps
☁️ Google Cloud Ready	Designed to run on Google Cloud Vertex AI / Compute Engine
🔌 Custom MCP Server	Built from scratch in TypeScript — 6 tools covering all LeetCode actions
⚡ Constraint-aware	Maps n-constraints to complexity ceilings before writing any code
🎯 Max 4 Retries	Never loops forever — documents failures and moves on

🏗️ Architecture

┌─────────────────────────────────────────────────────────────────────────┐
│                        YOUR MACHINE / GOOGLE CLOUD                      │
│                                                                         │
│  ┌──────────────────────────────────────────────────────────────────┐   │
│  │                     PYTHON AGENT  (agent.py)                     │   │
│  │                                                                  │   │
│  │   ┌─────────────┐    ┌──────────────┐    ┌──────────────────┐   │   │
│  │   │  Queue Mgr  │───▶│  Claude      │───▶│  Result Handler  │   │   │
│  │   │  Skip logic │    │  Opus 4      │    │  WA/TLE/RE/MLE   │   │   │
│  │   │  Retry loop │    │  (AI brain)  │    │  Retry / Move on │   │   │
│  │   └─────────────┘    └──────┬───────┘    └──────────────────┘   │   │
│  │                             │ MCP Protocol (stdio)               │   │
│  └─────────────────────────────┼────────────────────────────────────┘   │
│                                │                                         │
│  ┌─────────────────────────────▼────────────────────────────────────┐   │
│  │              LEETCODE MCP SERVER  (Node.js / TypeScript)         │   │
│  │                                                                  │   │
│  │  ┌────────────┐  ┌──────────────┐  ┌───────────┐  ┌─────────┐  │   │
│  │  │check_auth  │  │list_problems │  │get_problem│  │ submit  │  │   │
│  │  └────────────┘  └──────────────┘  └───────────┘  └─────────┘  │   │
│  │  ┌────────────┐  ┌──────────────┐                               │   │
│  │  │get_result  │  │get_progress  │                               │   │
│  │  └────────────┘  └──────────────┘                               │   │
│  └─────────────────────────────┬────────────────────────────────────┘   │
│                                │ HTTPS / GraphQL                         │
└────────────────────────────────┼────────────────────────────────────────┘
                                 │
                    ┌────────────▼────────────┐
                    │   leetcode.com API       │
                    │   GraphQL + REST         │
                    │   Your account session   │
                    └─────────────────────────┘

📁 Project Structure

AUTONOMOUS-LEETCODE-AGENT/
│
├── 📂 mcp-server/                  ← Custom MCP Server (TypeScript)
│   ├── 📂 src/
│   │   ├── 📄 index.ts             ← Server entry + tool registry
│   │   ├── 📄 leetcode-client.ts   ← All HTTP/GraphQL calls to LeetCode
│   │   └── 📂 tools/
│   │       ├── 📄 check_auth.ts    ← Verify session cookies
│   │       ├── 📄 get_problem.ts   ← Fetch full problem details
│   │       ├── 📄 get_progress.ts  ← Account solve statistics
│   │       ├── 📄 get_result.ts    ← Poll submission verdict
│   │       ├── 📄 list_problems.ts ← List/filter all problems
│   │       └── 📄 submit_code.ts   ← Submit solution
│   ├── 📄 package.json
│   └── 📄 tsconfig.json
│
├── 📂 agent/                       ← Autonomous Agent (Python)
│   ├── 📄 agent.py                 ← Main loop + Claude API calls
│   ├── 📄 requirements.txt
│   ├── 📄 .env.example             ← Config template
│   ├── 📄 .env                     ← Your credentials (never commit!)
│   └── 📂 logs/                    ← JSON session logs
│
├── 📄 run.sh                       ← One-click setup & run script
├── 📄 .gitignore
└── 📄 README.md                    ← You are here

⚡ Quickstart

Step 1 — Clone the repo

git clone https://github.com/Vellorpavan/AUTONOMOUS-LEETCODE-AGENT.git
cd AUTONOMOUS-LEETCODE-AGENT

Step 2 — Install everything

./run.sh setup

This installs Node.js dependencies, compiles the TypeScript MCP server, creates a Python virtual environment, and installs Python packages.

Step 3 — Get your LeetCode cookies Your agent authenticates using your browser session. LeetCode does not provide a public API key, so this is the standard approach.

Log into leetcode.com in Chrome or Firefox
Press F12 to open Developer Tools
Click the Application tab (Chrome) or Storage tab (Firefox)
In the left sidebar: Cookies → https://leetcode.com
Find and copy these two values:

Cookie Name	What it looks like
`LEETCODE_SESSION`	Very long string, 200+ chars, starts with `eyJ...`
`csrftoken`	Shorter string, ~32 chars

⚠️ These cookies expire every 2–4 weeks. If the agent stops authenticating, repeat this step to get fresh cookies.

Step 4 — Configure credentials

cp agent/.env.example agent/.env

Open agent/.env and fill in your values:

# Anthropic API — get from https://console.anthropic.com/settings/keys
ANTHROPIC_API_KEY=sk-ant-your-key-here

# LeetCode cookies — from your browser (see Step 3)
LEETCODE_SESSION=eyJ0eXAiOiJKV1Q...your-long-session-here
LEETCODE_CSRF_TOKEN=your-csrf-token-here

# Agent settings
DEFAULT_LANG=python3       # python3 | javascript | java | cpp
MAX_RETRIES=3              # attempts per problem before moving on

Step 5 — Run

./run.sh solve-all

That's it. The agent runs until every unsolved problem is attempted.

💻 Usage

# ── Core Commands ──────────────────────────────────────────────────

./run.sh solve-all          # Solve ALL unsolved problems (full grind)
./run.sh solve-easy         # Only Easy problems
./run.sh solve-medium       # Only Medium problems
./run.sh solve-hard         # Only Hard problems

# ── Specific Problem ───────────────────────────────────────────────

./run.sh solve two-sum
./run.sh solve longest-substring-without-repeating-characters
./run.sh solve median-of-two-sorted-arrays

# ── Advanced Options ───────────────────────────────────────────────

cd agent && source venv/bin/activate

python3 agent.py --lang javascript     # Use JavaScript
python3 agent.py --limit 50            # Stop after 50 problems
python3 agent.py --difficulty MEDIUM   # Only mediums
python3 agent.py --slug two-sum        # One specific problem

What you see while it runs

╔══════════════════════════════════════════════╗
║     LeetCode Autonomous Agent  🤖            ║
║  Language: python3  |  Max retries: 4        ║
╚══════════════════════════════════════════════╝

Session start. 847 solved, 2174 remaining.

━━━━━━━━━━ Problem: two-sum ━━━━━━━━━━
✅ #001 | Two Sum | Easy | Accepted | attempt 1 | O(n) hashmap | 52ms

━━━━━━━━━━ Problem: add-two-numbers ━━━━━━━━━━
✅ #002 | Add Two Numbers | Medium | Accepted | attempt 1 | O(n) linked list

━━━━━━━━━━ Problem: longest-substring... ━━━━━━━━━━
✅ #003 | Longest Substring | Medium | Accepted | attempt 1 | O(n) sliding window

...

══════════════════════════════════════
SESSION COMPLETE
✅ Accepted:   312 problems
❌ Failed:      14 problems
⏭  Skipped:    87 problems (premium)
⏩ Pre-solved: 847 problems
══════════════════════════════════════

🛠️ MCP Server — Tool Reference

The MCP server exposes 6 tools to the AI agent:

📋 check_auth — Verify your LeetCode session

check_auth()
// Returns: { isAuthenticated: true, username: "your_username" }
// Always called first. If false → agent stops immediately.

📊 get_progress — Your solve statistics

get_progress()
// Returns:
// {
//   total: 3279,
//   solved: 847,
//   tried: 124,
//   untouched: 2308,
//   byDifficulty: {
//     EASY:   { total: 882,  solved: 612 },
//     MEDIUM: { total: 1849, solved: 213 },
//     HARD:   { total: 548,  solved: 22  }
//   },
//   progressPercent: "25.8%"
// }

📃 list_problems — Fetch the problem queue

list_problems({
  unsolvedOnly: true,   // only unsolved
  fetchAll: true,       // all pages
  difficulty: "EASY",   // optional filter
})
// Returns array of { id, title, slug, difficulty, status, isPaidOnly }

🔍 get_problem — Full problem details

get_problem({ slug: "two-sum" })
// Returns:
// {
//   id: "1",
//   title: "Two Sum",
//   difficulty: "Easy",
//   description: "Given an array of integers...",
//   exampleTestcases: "...",
//   starterCode: { python3: "class Solution:\n    def twoSum...", ... },
//   hints: [...],
//   tags: ["Array", "Hash Table"]
// }

📤 submit_code — Submit a solution

submit_code({
  slug: "two-sum",
  lang: "python3",
  code: "class Solution:\n    def twoSum(self, nums, target):\n        ..."
})
// Returns: { submissionId: "1234567890", success: true }

✅ get_result — Poll the verdict

get_result({ submission_id: "1234567890" })
// Returns:
// {
//   verdict: "Accepted",
//   isAccepted: true,
//   runtime: "52ms",
//   memory: "16.4MB",
//   runtimePercentile: "Beats 94.2% of submissions",
//   testsPassed: "63/63"
// }
// OR on failure:
// {
//   verdict: "Wrong Answer",
//   failingTestcase: "[2,7,11,15]\n9",
//   expectedOutput: "[0,1]",
//   actualOutput: "[1,0]"
// }

🤖 Agent System Prompt

The agent uses an elite 5-part system prompt:

PART 1 — IDENTITY
  Elite competitive programmer. 10+ years ICPC + LeetCode.
  Never guesses. Verifies everything.

PART 2 — AGENT LOOP
  Startup → auth check → load queue → sort Easy/Medium/Hard
  Per problem: skip check → solve → submit → check → branch on verdict
  Rate limit guard (wait 60s on 429) → session log → next problem

PART 3 — SOLVER BRAIN (per problem)
  Constraint → complexity map (n≤20: O(2ⁿ), n≤10⁶: O(n)...)
  10-step workflow: read → pattern → algorithm → complexity check
  → edge cases → implement → dry run → validate → debug → output

PART 4 — HARD RULES
  Never invent test cases. Never submit same code twice.
  Never loop more than 4 times. Never wait for human input.
  One failed problem never stops the session.

PART 5 — LOG FORMAT
  ✅ #001 | Two Sum | Easy | Accepted | attempt 1 | O(n) hashmap
  ❌ #042 | Trapping Rain Water | Hard | Failed | attempt 4 | TLE

📊 Session Logs

Every run saves a full JSON log to agent/logs/:

{
  "session_id": "20240315_143022",
  "start_time": "2024-03-15T14:30:22",
  "duration_seconds": 14402,
  "total": 312,
  "accepted": 298,
  "failed": 9,
  "skipped": 5,
  "results": [
    {
      "timestamp": "2024-03-15T14:30:45",
      "id": "1",
      "title": "Two Sum",
      "slug": "two-sum",
      "difficulty": "Easy",
      "status": "accepted",
      "attempts": 1,
      "lang": "python3",
      "notes": "O(n) hashmap"
    },
    {
      "id": "42",
      "title": "Trapping Rain Water",
      "difficulty": "Hard",
      "status": "failed",
      "attempts": 4,
      "notes": "TLE on monotonic stack approach"
    }
  ]
}

☁️ Google Cloud Deployment

This project is built to run on Google Cloud for 24/7 autonomous operation.

Deploy to Google Cloud Compute Engine

# 1. Create a VM instance
gcloud compute instances create leetcode-agent \
  --machine-type=e2-standard-2 \
  --image-family=ubuntu-2204-lts \
  --image-project=ubuntu-os-cloud \
  --zone=us-central1-a

# 2. SSH into the instance
gcloud compute ssh leetcode-agent --zone=us-central1-a

# 3. Install dependencies on the VM
sudo apt update && sudo apt install -y nodejs npm python3 python3-pip python3-venv git

# 4. Clone your repo
git clone https://github.com/Vellorpavan/AUTONOMOUS-LEETCODE-AGENT.git
cd AUTONOMOUS-LEETCODE-AGENT

# 5. Setup
./run.sh setup

# 6. Add credentials
nano agent/.env

# 7. Run in background (stays alive after SSH disconnect)
nohup ./run.sh solve-all > agent/logs/stdout.log 2>&1 &

# 8. Watch it run
tail -f agent/logs/stdout.log

Deploy with Google Cloud Run (Containerized)

# Dockerfile (add to project root)
FROM node:18-slim AS mcp-builder
WORKDIR /app/mcp-server
COPY mcp-server/package*.json ./
RUN npm install
COPY mcp-server/ ./
RUN npm run build

FROM python:3.11-slim
WORKDIR /app
RUN apt-get update && apt-get install -y nodejs npm
COPY --from=mcp-builder /app/mcp-server/dist ./mcp-server/dist
COPY --from=mcp-builder /app/mcp-server/node_modules ./mcp-server/node_modules
COPY agent/requirements.txt ./agent/
RUN pip install -r agent/requirements.txt
COPY agent/ ./agent/
CMD ["python3", "agent/agent.py"]

# Build and push
gcloud builds submit --tag gcr.io/YOUR_PROJECT/leetcode-agent

# Deploy
gcloud run deploy leetcode-agent \
  --image gcr.io/YOUR_PROJECT/leetcode-agent \
  --platform managed \
  --set-env-vars ANTHROPIC_API_KEY=...,LEETCODE_SESSION=...,LEETCODE_CSRF_TOKEN=...

🐛 Troubleshooting

❌ "Not authenticated" error

Your cookies have expired (they last 2–4 weeks).

Log into leetcode.com in your browser
Get fresh cookies (see Getting Your LeetCode Cookies)
Update agent/.env with the new values
Restart the agent

❌ "MCP server not found" / "Cannot find module"

The TypeScript hasn't been compiled yet.

cd mcp-server
npm install
npm run build

❌ Agent stops after one problem

Check agent/logs/ for the latest session file. Look for the error. Common causes:

API rate limit hit → add time.sleep(2) between problems
Network timeout → check internet connection
Wrong MCP path → check MCP_SERVER_PATH in .env

❌ "Premium problem" errors

The agent automatically skips premium-only problems. If you have LeetCode Premium, the cookies will already include premium access — no extra config needed.

❌ Python venv issues

cd agent
rm -rf venv
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

🗺️ Roadmap

🤝 Contributing

Project Details: This is a real, working project developed with significant engineering hours to autonomously solve problems on LeetCode. (Note: This project is proprietary and does not have an open-source license.)

Contributions are welcome! Here's how:

# Fork the repo, then:
git clone https://github.com/Vellorpavan/AUTONOMOUS-LEETCODE-AGENT.git
cd AUTONOMOUS-LEETCODE-AGENT

# Create a feature branch
git checkout -b feature/your-feature-name

# Make your changes, then
git commit -m "feat: your feature description"
git push origin feature/your-feature-name

# Open a Pull Request on GitHub

Built with 🤖 Claude Opus 4 · ☁️ Google Cloud · 🔌 MCP Protocol

If this project helped you — drop a ⭐ on GitHub. It means everything.

Navigation Menu

Repository files navigation

📺 [Click to Watch] Autonomous LeetCode Agent in Action!

📌 Table of Contents

🧠 What Is This?

✨ Features

🏗️ Architecture

📁 Project Structure

⚡ Quickstart

💻 Usage

What you see while it runs

🛠️ MCP Server — Tool Reference

🤖 Agent System Prompt

📊 Session Logs

☁️ Google Cloud Deployment

🐛 Troubleshooting

🗺️ Roadmap

🤝 Contributing

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
agent		agent
mcp-server		mcp-server
.gitignore		.gitignore
README.md		README.md
run.sh		run.sh

Latest commit

History

Folders and files

Repository files navigation

📺 [Click to Watch] Autonomous LeetCode Agent in Action!

📌 Table of Contents

🧠 What Is This?

✨ Features

🏗️ Architecture

📁 Project Structure

⚡ Quickstart

💻 Usage

What you see while it runs

🛠️ MCP Server — Tool Reference

🤖 Agent System Prompt

📊 Session Logs

☁️ Google Cloud Deployment

🐛 Troubleshooting

🗺️ Roadmap

🤝 Contributing

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages