Skip to content
New Scheduled tasks: run a browser job once or on repeat

An agentic browser harness
for the agent CLI you already use

Browsentic connects Claude Code, Codex, Cursor or another agent CLI to your real, signed-in browser. You type or say the job in the side panel. The agent works in your tab and asks before it submits, uploads or leaves the site.

For example: every weekday at 9, summarise the PRs waiting on me

One line, about two minutes

On macOS

curl -fsSL https://browsentic.com/install.sh | sh

On Windows, in PowerShell (experimental)

irm https://browsentic.com/install.ps1 | iex

Anywhere, with Node.js 20 or newer

npx browsentic@latest setup

Setup asks which browser you use and opens the extension’s store page there. Press Add, click Browsentic in the toolbar and enter the code setup shows, once. Install guide

Runs on your machine

nothing hosted, no cloud browser

No API key

it runs the agent you are already signed in to

Your logged-in session

your cookies, your two-factor, your extensions

Free and open source

Apache 2.0 licensed, no account, no subscription

What it does

What you can hand the agent in your own browser

Scheduled jobs, recorded jobs, several tabs at once, site maps and voice, all on the sites you are already signed in to.

New

Put any browser job on a schedule

Every weekday at nine, every Friday evening, once tomorrow morning. Each run works in a background tab of its own and leaves you a one-line result. A recorded job replays without an agent, so it spends no tokens.

Scheduled tasks →

Record a job once, replay it later

Press record and do the job yourself. Next time, “do it like last time” is the whole instruction, and every step is checked against the live page before it runs.

Record and replay →

Several tabs working at once

Start a job in one tab, then another. Each keeps its own conversation and keeps working in the background while you read something else.

Orchestration →

Map a web app before working in it

Point the agent at a web app and it explores the site, then saves notes that every later visit to that site starts from.

Site maps →

Give instructions by voice

Close the panel and speak. A mic floats on the page, and a short pause sends what you said. In Chrome and Edge.

Hands-free →

How it works

Everything runs on your own computer

The side panel passes your instruction to Browsentic Bridge, a small program on your computer, which starts the agent CLI you already use. The agent talks to its own model provider as usual. Browsentic adds no server.

Your computer

  1. You Type it, say it, or show it once
  2. Side panel In the browser you already use
  3. Browsentic Bridge A small local program that links the two
  4. Your agent Claude Code, Codex and more, already signed in

Safety

Safeguards for an agent in your real accounts

An agent in your browser can reach your email, your bank and your work. Browsentic connects to nothing until you pair it, and the agent must ask before it submits, uploads or leaves the site.

  • Off until you pair it

    A fresh install talks to nothing. You connect your browser yourself, with a one-time code.

  • Asks before it acts

    Sending a form, uploading a file or leaving the site it was sent to waits for your OK.

  • No server in between

    Your agent talks to its own provider, as it always does. Browsentic has no cloud of its own.

  • Open source, Apache 2.0

    Read every line. Code the agent writes for itself is shown to you in full before it runs.

Two limits. Other programs running as you could drive a browser you have already paired. A hostile page can still try to mislead the agent, as it could any agent that reads the web.

FAQ

Frequently asked questions

How do I install it?

Run one line. On a Mac, curl -fsSL https://browsentic.com/install.sh | sh installs the Mac app; on Windows, irm https://browsentic.com/install.ps1 | iex installs the Windows app (experimental); anywhere with Node.js 20 or newer, run npx browsentic@latest setup. Each installs Browsentic Bridge on your computer, asks which browser you use and opens the extension’s store page there (the signed add-on for Firefox). Add it, click Browsentic in the toolbar and enter the code it shows, once. Both parts update themselves from then on.

Do I need an API key?

No. Browsentic runs on the agent CLI login you already have: Claude Code, Codex, Antigravity, Mistral Vibe, Grok Build, Cursor CLI, Qwen Code or OpenCode. There is no API client in the repository and nothing to paste into a settings field. Browsentic Bridge starts your CLI locally, as you.

Can it run a job on a schedule?

Yes. In the side panel’s Schedules tab, set an instruction or a recording to run once at a set time, on chosen days, or every so often. Each run opens its own background tab, does the job, closes the tab and leaves a one-line result on the page you are looking at. A recording replays without an agent and spends no tokens, unless a step no longer fits the page. Runs need the browser open; one that falls due while it is closed runs once when the browser is back. Anything that needs your approval shows a card on the page you are on.

Is this a headless browser?

No. It drives the real tab in front of you, with your sessions, cookies, extensions and two-factor state. Nothing is re-authenticated in a throwaway profile.

Is Browsentic an MCP server or a browser extension?

Neither on its own. Browsentic is a browser harness with three parts: a side panel where you give your agent work and approve what it does, the extension that carries that panel (from a browser store, or a signed add-on for Firefox), and Browsentic Bridge on your computer, which starts your agent CLI. The Bridge also serves an optional MCP endpoint, so Claude Code, Cursor or Zed can drive the same browser from a terminal. The side panel does not depend on it.

Give your browser a superpower.

Free and open source under Apache 2.0, with no API key, subscription or account. One line installs Browsentic Bridge and opens the extension’s store page. Nothing connects until you pair the two.