Browser Use Terminal Turns Your Command Line Into a Full Browser Agent
Browser Use Terminal is a Rust-built TUI that lets you run browser agents from the command line, controlling real Chrome with a new LLM harness claimed to be 2x cheaper and faster.
Takeaways
- Browser Use Terminal is a new open-source Rust TUI that runs browser agents from the CLI with a one-line install.
- It connects to your real Chrome session, letting agents act on sites where you're already logged in — no credential setup needed.
- Built on a new Rust LLM harness claimed to be 2x cheaper and faster than the team's previous Browser Harness approach.
- Supports three browser modes: real Chrome, headless Chromium, and Browser Use cloud — switchable via
/browsercommand. - Integrates with Codex auth for OpenAI users and supports multiple LLMs via a
/modelslash command. - Free and MIT-licensed; available now at github.com/browser-use/terminal at v0.1.1.
Browser Use Terminal just shipped: a command-line interface that turns your terminal into a full browser agent. Type a task, watch Chrome execute it. No Python script, no GUI, no boilerplate. Just one line to install and a prompt to run.
The project comes from the same team behind the popular browser-use open-source library, which has accumulated nearly 96,000 GitHub stars. The terminal tool is their bet that the best interface for browser automation is not a web dashboard or a Python SDK, but the terminal itself.
One curl, then you're running
Installation is a single command:
curl -fsSL https://browser-use.com/terminal/install.sh | sh
browser
Browser Use Terminal is a Rust TUI (terminal user interface) for browser agents, combining a new LLM harness, Browser Harness-style CDP control, real Chrome sessions, and a terminal UI you can actually steer. Once running, you control it through slash commands:
/auth, sign in (supports Codex auth)/model, choose your LLM/browser, pick local Chrome, headless Chromium, or cloud/update, update the app in place

The tool is free and open-source under the MIT license. Running it costs whatever your LLM API calls cost. The terminal itself has no subscription fee.
What's actually under the hood
The architecture is more interesting than the demo suggests. Browser Use Terminal is a browser-first LLM harness: Rust owns the agent loop and durable state, while the browser runtime gives the model direct CDP control over Chrome.
CDP (Chrome DevTools Protocol) is the low-level wire protocol Chrome exposes for programmatic control. Most browser automation tools (Playwright, Puppeteer) wrap it in high-level abstractions. Browser Use Terminal hands it directly to the model, which means the agent can do anything Chrome can do, including things that higher-level wrappers deliberately prevent.
The internal stack looks like this:
- Ratatui UI , a custom Rust terminal renderer for live task watching, steering, stopping, and resuming
- Rust LLM harness , manages tools, subagents, context compaction, and cancellation
- SQLite event log , persists history, screenshots, artifacts, and traces locally
- CDP browser runtime , handles Chrome profiles, crash recovery, and browser ownership
The model gets raw browser capability: CDP, page JavaScript, screenshots, files, and helper code. The team claims this new harness is 2x cheaper and 2x faster than their previous Browser Harness approach, though independent benchmarks are not yet available.
Three browser modes
The tool supports three browser modes: managed headless Chromium, real Chrome with existing user profiles and logins, and cloud-hosted browsers via the Browser Use Cloud API. The real Chrome mode is the most compelling. Because it connects to your actual logged-in browser session, the agent can access sites where you're already authenticated , your email, your admin dashboards, your SaaS tools , without you handing over credentials or setting up OAuth flows.
The README gives a sense of the intended task range:
- Get my San Francisco parking permit.
- Give this employee admin permission in Azure.
- Find the cancellation policy for my current hotel reservation.
These are tasks that require navigating real, logged-in web sessions , exactly the kind of thing that breaks when you try to automate them with a fresh headless browser that has no cookies.
The Codex connection
One notable feature is native Codex auth integration. Codex CLI is OpenAI's coding agent that you can run locally from your terminal. It can read, change, and run code on your machine in the selected directory. It's open source and built in Rust for speed and efficiency. Browser Use Terminal plugging into Codex auth means users already in the OpenAI ecosystem can authenticate without a separate API key setup , a small but meaningful reduction in friction.
The broader context here is that Rust-based terminal agents are becoming a pattern. Both Codex CLI and Browser Use Terminal are written in Rust, use SQLite for local state, and expose TUI interfaces built with Ratatui. The design philosophy is converging: fast, durable, local-first agent loops with live terminal UIs.
What it's good at, and where it falls short
The tool shines for tasks that require real browser sessions with existing login state. Browser Use 1.0 is a purpose-built agent with stealth browser infrastructure, CAPTCHA solving, persistent filesystem, and optimized tool orchestration. The terminal interface adds the ability to watch, steer, stop, and resume tasks mid-execution , which matters a lot when an agent is about to click something irreversible.
The limitations are real though. The project is at v0.1.1, which means rough edges are expected. Tasks that require precise, multi-step form interactions on complex SPAs can still trip up browser agents. For browser automation specifically, the "best model" answer changes depending on the task complexity. Simpler navigation flows can run fine on cheaper models, but anything with dynamic content or multi-step decisions usually needs something stronger. And because the agent has raw CDP access, a misunderstood instruction can take real actions in real accounts , there's no sandbox by default when using your actual Chrome profile.
Why this matters beyond the demo
The deeper shift here is about where browser automation lives in the developer workflow. Until now, running a browser agent meant writing Python, setting up a virtual environment, managing API keys, and wiring up a script. Browser Use Terminal collapses that to a single install and a natural language prompt.
This positions browser automation as a terminal-native primitive , something you reach for the way you reach for curl or jq. A persistent background daemon keeps the browser alive between commands for fast ~50ms latency. Combined with the SQLite-backed history and artifact storage, it starts to look less like a demo tool and more like infrastructure.
The repo is public, MIT-licensed, and actively developed with 310 commits already on the main branch. For anyone who has wanted to automate browser tasks without leaving the terminal, this is the most complete starting point available today.