Unbrowse — The API layer for AI agents
Unbrowse is an open-source CLI tool that reverse-engineers the internal APIs behind any website, allowing AI agents to make direct API calls instead of automating headless browsers. It reduces page interaction time from 5-30 seconds to 50-200 milliseconds and cuts token usage from ~8,000 to ~200 tokens per action. Skills discovered by one agent are shared in a public registry for all agents to reuse.
Setup: npx unbrowse setup
Upgrade existing installs: npm install -g unbrowse@latest && unbrowse setup
For skill hosts: npx skills add unbrowse-ai/unbrowse
Community: https://discord.gg/VWugEeFNsG
Usage: unbrowse resolve --intent "..." --url "..."
Full documentation: https://www.unbrowse.ai/skill.md
100x faster. 95% cheaper.
The API-native browser.
A drop-in replacement for browser automation for AI agents. Log in, search, book, and submit through direct API calls instead of driving a flaky browser.
Install First
Install Unbrowse and start using it.
Fresh install or quick upgrade. Same setup flow either way.
# Full setup in one commandnpx unbrowse setup# Install globally for repeat usenpm install -g unbrowseunbrowse setup# Already installed? Upgrade to latest after releasesnpm install -g unbrowse@latestunbrowse setup# Add the skill for agent workflowsnpx skills add unbrowse-ai/unbrowse# Use itunbrowse resolve --intent "get events" --url "https://lu.ma"
npm install -g unbrowse@latestunbrowse setupWorks seamlessly with
What is Unbrowse?
Unbrowse is an open-source drop-in replacement for browser automation when you want API-native access to the web. It reverse-engineers the internal APIs behind any website so AI agents can make direct API calls instead of driving headless browsers. It reduces page interaction time from 5–30 seconds to 50–200 milliseconds and cuts token usage from ~8,000 to ~200 tokens per action. Skills discovered by one agent are shared in a public registry for all agents to reuse.
Three ways to see the same website
What Humans See
Beautiful, interactive UI
What Agents See Today
Image + 847KB DOM × 1 passes
What Unbrowse Does
1KB JSON × 0 calls
Bypass the DOM completely.
Skip the rendering engine
Headless browsers are slow and flaky. Unbrowse taps directly into the hidden internal APIs that power the frontend, returning data instantly.
40x fewer tokens
Why burn context on 8,000 tokens of HTML? Your agent gets the exact JSON data it needs to take the next action — nothing else.
Auto-discovers APIs
Your agent types unbrowse resolve --intent "..." --url "...". We instantly map the site's undocumented endpoints for immediate use.
Integrate with anything. Behind auth.
No cloud proxies, no expensive credits. Unbrowse runs locally, leveraging your actual browser sessions to securely access auth-protected content.
One agent discovers it.
Every agent benefits.
Every time an agent maps a new website, the underlying API endpoints are published to a shared registry. Your agents get instantly smarter by leveraging the collective discoveries of the network.
Recently Indexed Skills
View full registryLoading registry skills...
Works with your stack
Example: airbnb.com
One agent browses Airbnb. Every agent on the network can now search listings, check availability, and book — instantly, no browser.
Register, verify, and start using it.
Once the CLI is installed, grab your key, verify the local server, and jump into the docs and community.
Frequently Asked Questions
How does Unbrowse work?
Unbrowse is a drop-in replacement for browser automation when you want API-native access to the web. It opens a local browser, captures network traffic as you interact with a site, and reverse-engineers the internal API endpoints that power the frontend. Once discovered, these endpoints are stored as reusable skills so your agent can call them directly — no browser required.
How much faster is Unbrowse than headless browser automation?
Unbrowse is approximately 100x faster per page. Headless browsers typically take 5–30 seconds per page interaction. Unbrowse makes direct API calls in 50–200 milliseconds. It also uses ~200 tokens per action compared to ~8,000 tokens for scraped HTML, a 40x reduction.
Is Unbrowse free?
Yes. Unbrowse is 100% free and open source under the AGPL-3.0 license. There are no paid tiers, cloud proxies, or usage credits. Everything runs locally on your machine.
What websites does Unbrowse support?
Unbrowse works with any website that uses internal APIs to power its frontend — which includes most modern web applications. Sites like Airbnb, LinkedIn, and hundreds of others have been successfully mapped. When a site cannot be reverse-engineered, Unbrowse falls back to standard browser automation.
Is Unbrowse secure? Do my credentials leave my machine?
Unbrowse runs entirely locally. There are no cloud proxies, no man-in-the-middle interception, and your browser cookies never leave your device. Authentication credentials are encrypted with AES-256-CBC in a local vault. Only discovered API endpoint patterns (not data or credentials) are shared with the registry.
How do I install Unbrowse?
Run npx unbrowse setup for a one-command installation that sets up browser assets and configures your agent host. If Unbrowse is already installed, upgrade to the latest version with npm install -g unbrowse@latest and rerun unbrowse setup. For skill-based agent platforms like OpenClaw, use npx skills add unbrowse-ai/unbrowse.
What is the skill registry?
The skill registry is a shared marketplace of reverse-engineered API skills. When one agent discovers how to interact with a website's API, that knowledge is published to the registry so every other agent can use it instantly — no need to re-discover the same endpoints.