HomeToolsMCPHow It WorksStoriesPhilosophyArchitectureStar on GitHub
All Tools
L
AgentsFreeOpen Source

LIGHTFEED BROWSER AGENT

AI drives the browser once, then replay navigation with zero tokens

MIT

ABOUT

Browser automation agents burn LLM tokens on every run, even when navigation steps are identical across executions. Most token spend goes to repetitive clicking, scrolling, and typing to reach a target page, while only the final data extraction needs fresh AI reasoning. Lightfeed Browser Agent splits this into two phases: AI navigation that can be recorded and replayed with zero tokens, and cheap extraction that runs only on the result page. Optional aiFallback re-plans only broken steps when the DOM drifts, cutting token waste to a fraction.

INSTALL
npm install @lightfeed/browser-agent

INTEGRATION GUIDE

1. Schedule recurring web scraping jobs that navigate to the same pages daily without burning tokens on navigation 2. Build a data extraction pipeline that replays recorded browser paths and only runs AI for extracting live content 3. Run browser agents on AWS Lambda with the serverless Chromium provider for cost-effective scheduled scraping 4. Create deterministic browser test suites that replay navigation plans for consistent, fast UI testing 5. Prototype browser agents locally with Chromium, then deploy the same code to a remote browser farm via CDP endpoint

TAGS

browser-automationweb-scrapingplaywrighttoken-optimizationserverlesschromiumtypescriptextraction