omxyz

omxyz / lumen

Public

Vision-first browser agent

57
0
100% credibility
Found Mar 07, 2026 at 57 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Lumen is a library for building AI agents that autonomously navigate websites using natural language instructions, screenshots, and vision models.

How It Works

1
🔍 Discover Lumen

You hear about Lumen, a helpful tool that lets AI browse websites and complete tasks just by telling it what to do in simple words.

2
📦 Set it up quickly

You add Lumen to your project with a simple download, and everything is ready to go in moments.

3
🧠 Connect a smart AI

You link it to a thinking AI service like Claude so it can understand screenshots and decide what to do next.

4
Choose your browser style
💻
Local browser

Run it right on your computer to see the action live.

☁️
Cloud browser

Use a secure online browser that works anywhere without setup.

5
💬 Give it a task

You type something everyday like 'Find the top story on Hacker News and tell me the title' – it springs to life!

6
👀 Watch it work

It takes screenshots, clicks buttons, scrolls pages, and figures things out step by step, just like you would.

✅ Get your answer

It finishes the job and hands you the exact result, saving you time on boring web hunts.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 57 to 57 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is lumen?

Lumen (omxyz/lumen on GitHub) is a TypeScript Node.js library for vision-first browser agents. Give it plain-English tasks like "find the top Hacker News story," pair with a vision model (Claude Sonnet, Gemini, GPT), and it drives Chrome via screenshots and pixel actions—no DOM selectors or scraping. Handles real browsers locally or via Browserbase, with session resumption for long runs.

Why is it gaining traction?

Crushes benchmarks: 96% WebVoyager pass rate (24/25 tasks), 32% token savings over baselines via screenshot compression and LLM summaries. Streaming events build real-time UIs, safety policies block risky domains/actions, repeat detection nudges stuck loops. Multi-provider (Anthropic/Google/OpenAI) and action caching speed iteration without brittle hacks.

Who should use this?

AI devs automating web tasks: scrape arXiv prices, book Google Flights, or extract GitHub stars. Node teams replacing flaky Playwright/Selenium scripts with LLM-driven agents. Prototypers testing browser-use models on live sites like Allrecipes or Wikipedia.

Verdict

Strong evals make this lumen ai github agent worth a spin for prototypes, but 55 stars and 1.0% credibility score reflect early maturity—solid docs/tests, but scale cautiously. (187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.