A real browser for AI agents.

01 / Start building

# Save a rendered page to a file
rel https://example.com > example.html

# Navigate, wait, and capture the current page
rel navigate https://example.com
rel perform '[{"action":"wait-for","selector":"body"}]'
rel capture > page.html

from rel_playwright.sync_api import sync_playwright

with sync_playwright() as playwright:
    browser = playwright.chromium.launch()
    page = browser.new_page()
    page.goto("https://example.com")
    print(page.title())
    browser.close()

{
  "mcpServers": {
    "rel": {
      "command": "/Applications/REL.app/Contents/Resources/rel-mcp"
    }
  }
}

02 / The browser layer, handled

1. 01

   Stay in the same browser ↗

   Watch an agent work, step in when it needs you, and hand control back. Your browser sessions stay visible on your Mac.

2. 02

   Keep your sessions separate ↗

   Each session has its own cookies, storage, and browser identity. Keep saved logins and return to the work later.

3. 03

   Act through Chromium ↗

   Click, type, scroll, and navigate with native browser input. Agents use the same controls you see on the page.

4. 04

   Give every session its own route ↗

   Use a different proxy for each session, including authenticated proxies. Pause a session to cut its network access instantly.

5. 05

   Make the page useful to a model ↗

   Read rendered HTML or Markdown, inspect semantic structure, capture pixels, or combine structure and screenshots in one observation.

6. 06

   Save a setup. Use it again. ↗

   Reuse profiles with your proxy, browser identity, and filtering choices. Set exact viewports for repeatable captures and tests.

7. 07

   Spend less on proxy traffic ↗

   Block ads, trackers, or heavy images with per-session bandwidth controls. Keep more of your metered traffic for the content you need.

8. 08

   Put recurring work on a schedule ↗

   Choose a profile and model, write a prompt, and set a schedule. Give recurring research and browser tasks a place to run.

03 / Measured in a real browser

Tested in RELDebug on macOS with 50 navigation samples per stage. These are development-build results, not a certification of the current download. Chromium 152 validation is pending.