Capture browsing sessions and index callable endpoints as reusable API routes.
npx skills add https://github.com/barkleesanders/claude-code-starter --skill unbrowse
Please help me install this Agent Skill. Skill: unbrowse Source: https://github.com/barkleesanders/claude-code-starter/tree/main/skills/unbrowse Command: npx skills add https://github.com/barkleesanders/claude-code-starter --skill unbrowse
Agents waste time and tokens repeatedly browsing the same websites with generic browser automation. Unbrowse captures a site once and turns it into a reusable, indexed API route that future agents can call directly — typically 30× faster and 90% cheaper than a fresh browser session.
Use the unbrowse skill to resolve the intent "search hacker news for openai" against https://news.ycombinator.com and execute the returned endpoint to get structured results.
High-intent search queries and answers about installing and using this skill.
To turn websites into API routes for AI agents, you capture a live browsing session once and index the callable endpoints. Unbrowse transforms web platforms into stable, schema-validated API calls replacing brittle DOM scraping.
Capturing a browsing session replaces web scraping by recording website interactions once and indexing them as reusable API routes. This mechanism avoids brittle DOM selectors, providing stable endpoints with schema validation for future agent calls.
Unbrowse supports Playwright and Firecrawl by providing drop-in shims for these browser automation tools. It also offers an MCP server, CLI, and TypeScript SDK, allowing the same runtime to access captured routes across multiple surfaces.
The best way to reduce tokens spent on repetitive agent browser sessions is to capture the site once and reuse the indexed API route. Route replay executes discovered endpoints directly, running significantly faster and cheaper than fresh browser sessions.
You need a local runtime to execute unbrowse, along with optional MCP host integration for agent workflows. Marketplace connectivity is also required for discovering existing routes and publishing new ones captured from your browsing sessions.
You should not use reusable API routes for web automation when interacting with highly dynamic or one-off web pages that change structure unpredictably. Unbrowse is designed for repeated agent workflows on stable platforms like search engines, SaaS tools, and e-commerce.
hhopkins95Configure a Node.js agent-runtime backend with REST and WebSocket servers.
browser-actRetrieve official websites and social profiles from a company name or website URL.
ryanbreenAutomate interactive GDB debugging sessions for the Breenix kernel.
J-StaR-Films-StudiosExtract captions, comments, and engagement metrics from social media via browser automation.
