How it works Skills Connections Providers Add to Chrome — Free
Public Beta — v1.0.3

Tell your browser
what to do.

Automate work in Chrome with plain English — BrowseAgent clicks, reads, fills forms, and navigates for you.

Add to Chrome — Free View on GitHub →
How It Works

One task. Three clear states.

Write the goal, watch the run, keep the result. The workflow stays visible inside Chrome from start to finish.

01

Start with a prompt

Describe the outcome you want. BrowseAgent turns it into browser actions.

Prompt state
Prompt entered in BrowseAgent side panel


02

Watch each step

The panel shows what the agent is doing while it runs, so the flow stays understandable.

Execution state
BrowseAgent showing execution steps during a task


03

Get the result

End with a usable answer, send it to a tool, or keep going from the same panel.

Result state
BrowseAgent showing a completed result summary
Skills

Built for real browser tasks.

Built to read pages, take action, send output, and avoid getting stuck in loops.

Browsing & Navigation

Navigate the web like an operator

Move across pages, tabs, and frames without losing context.

Example: open a search result, jump between tabs, then resume from a saved checkpoint.

Page Interaction

Work through pages, not demos

Click, type, scroll, and wait through real websites, not just static demos.

Example: fill a multi-step form or drive a research flow across dynamic pages.

Page Reading & Inspection

Read pages with structure

Read what is on screen with structured page understanding, search, and extraction.

Example: summarize a SERP, extract a pricing table, or find the right button from natural language.

Automation & Workflow

Queue, approve, resume

Approve plans, queue longer runs, and recover without starting over.

Example: approve the plan first, close the panel, then come back to the finished task.

Safety & Permissions

Guardrails built into the loop

Guardrails keep the agent from drifting into bad pages, repeated actions, or risky moves.

Example: pause on sensitive actions, reject duplicate calls, and ask before running scripts on a new domain.

External Integrations

Ship output where it belongs

Push results into the tools you already use instead of copying them by hand.

Example: post updates to Slack, write rows to Sheets, or send a final brief to Notion.

Tools

25 tools. One JSON schema.

A focused toolset for reading pages, taking action, and finishing work inside the browser.

Reading
6 tools

See the page, search it, and pull out the parts that matter.

read_page get_page_text extract_structured find_text
Navigation
9 tools

Move through sites, tabs, and frames without breaking the flow.

navigate open_tab switch_frame restore_snapshot
Interaction
7 tools

Take the actions a real operator would take to move the task forward.

click type press_key wait_for
External & Flow
5 tools

Call APIs, route output, save progress, and finish runs cleanly.

http_request notify_connector save_progress done
Connections

Inside the side panel.

Everything stays inside the side panel: skills, destinations, and provider setup.

BrowseAgent side panel overview with skills, connections, and provider setup
Providers

Bring your own model.

Use a hosted provider or run locally. The extension talks to the model directly from your browser.

◎
Z.AI API — GLM-4.6V
Recommended hosted option for vision and tool use.
See provider rates Recommended
✖
xAI — Grok 4.1 Fast
Fast hosted option with tool use. grok-4-1-fast-non-reasoning
See xAI pricing Budget
🦙
Ollama — Local Models
Run local models with no API key. Default: qwen3-vl:8b.
Free Free
Get started

Ready to automate
your browser?

Install BrowseAgent, connect a provider, and run your first real browser task in a couple of minutes.

Local-first. Open source. Built for operator-style browser work.

Add to Chrome View Source on GitHub