01
Simple capture
Open a URL in Chromium and save the rendered page.
rel https://example.com > example.htmlA real browser for AI agents.
REL is a native Mac browser built for AI agents, scripts, and scheduled tasks. Work in visible Chromium tabs, route sessions through different proxies, and automate with native browser input without page injection or DevTools Protocol.
- 01 Real browser tabs
- Keep agent work visible, organized, and ready to take over.
- 02 Human + agent control
- Watch, guide, or continue the work in the same Chromium window.
- 03 Native browser actions
- Click, type, and navigate without injected scripts.
- 04 Six ways to see a page
- Give models clean text, semantic structure, pixels, or both.
The browser layer, handled.
Open a tab, pick a profile or proxy, and start. Control the same visible browser from Chat, MCP, Python, the CLI, or REL’s local API.
One browser session per tab
Keep separate browser work visible, organized, and easy to revisit.
Native browser actions
Click, type, wait, and navigate through Chromium itself, from the CLI, MCP, RPC, or Rust.
Reusable profiles
Save your proxy, filtering, and browser setup once. Launch new sessions from it anytime.
Pause / Play
Cut a tab’s network access instantly, then reconnect when you’re ready.
Per-session proxies
Give each session its own route, including authenticated proxies.
Bandwidth controls
Block ads, trackers, or heavy images to spend less on metered proxies.
Custom browser identity
Shape Chromium’s identity for the work at hand without patching the page.
A terminal in every session
Open a ready-to-use terminal from any tab and start scripting.
Exact viewports
Pin desktop or mobile sizes for repeatable captures and tests.
Scheduled agent tasks
Choose a profile and model, set a schedule, and let REL run the job.
Automate from the CLI.
Capture a URL with one command, or keep a session open for multi-step work. Page data goes to standard output, so REL fits cleanly into scripts and pipelines.
02
Navigate, act, and capture
Keep one session open across a sequence of commands.
rel session create --name Research
rel navigate https://rel.me
rel perform '[{"action":"wait-for","selector":"body"}]'
rel capture > rel.htmlUse REL your way.
Start from the terminal, connect an agent, crawl with Python, or build directly on REL's local API.
CLI
Capture pages and automate browser work from the terminal.
Python crawler
Run long crawls in a real browser.
Playwright for Python
Bring your Playwright-shaped scraping code.
MCP server
Give agents a visible browser they can control.
RPC v1
Connect any local tool over HTTP.
Rust SDK
Build fast, typed integrations.
Choose what MCP returns.
Choose the view that fits the job: Markdown for reading, semantic elements for action, or synchronized pixels for visual pages. Every action stays grounded in the page state the model saw.
| Representation | What the model receives | Best for |
|---|---|---|
HTMLRendered HTML | Live rendered markup as a downloadable file. | Capture and debugging. |
MDSemantic Markdown | Focused page text and links. | Research and reading. |
SEMSemantic observation | Roles, states, bounds, and typed element references. | Fast, grounded interaction. |
SEM + PNGHybrid observation | A semantic view with its matching viewport image. | Interaction with visual context. |
PNGVisual observation | A viewport image with lightweight page context. | Canvas, charts, and maps. |
PNG / JPEG / WEBPStandalone screenshot | A viewport or full-page image, inline or saved. | Visual checks and evidence. |