Autonomous browser agent · Local LLM

Set the goal.
We browse.

and pause when it matters...

Browsie is a Chrome side-panel agent that reads the page like a human does, decomposes your goal into subgoals, then navigates, fills, clicks, and extracts until it's done. Runs against your local LLM by default — your data never leaves the box.

Local-first · LMStudio / Ollama · No telemetry · MV3 side panel · Free
NAVIGATE · FILL · CLICK · SCROLL · EXTRACT · WAIT_FOR · ASK_HUMAN · SET_FOCUS · DONE · NAVIGATE · FILL · CLICK · SCROLL · EXTRACT · WAIT_FOR · ASK_HUMAN · SET_FOCUS · DONE ·
§ 01 · Three roles, one goal

A planner, a doer, and a skeptic.

Most "browser agents" are one big prompt that loops until it hallucinates. Browsie isn't. Three small roles — the Director, the Actor, the Reflector — pass a checklist between them, and the agent itself stops to ask when it's unsure or auto-pauses when it's spinning. You can halt the run at any time.

// step 01

Director plans.

Reads your goal once and decomposes it into 1–6 ordered subgoals, each with a clear "done when" predicate. Runs once at the start of the run, not every step.

→ One LLM call
// step 02

Actor acts.

Each turn: snapshot the active tab as a compressed accessibility tree, pick exactly one tool action — fill, click, extract, navigate — and hand it to the executor. No DOM hallucination.

→ One action per turn
// step 03

Reflector judges.

After every action, the Reflector decides whether the current subgoal is satisfied. It advances the checklist, or it says "not yet" and the loop continues. When all subgoals are done, the run ends.

→ Honest "no" allowed
§ 02 · The toolkit

Ten verbs. No magic.

Browsie's actor only emits one of these tool calls per turn. That's the entire surface area. No code execution, no file write, no shell.

→ verb 01
navigate
Open a URL or follow a link in the current tab.
→ verb 02
fill
Type into a labelled input. Real keystroke events.
→ verb 03
click
Click a node by its accessible label or index.
→ verb 04
scroll
Move the viewport to reveal more nodes.
→ verb 05
extract
Pull a single value from the page.
→ verb 06
extract_records
Walk a list and return a typed table.
→ verb 07
wait_for
Block until a node or text appears.
→ verb 08
set_focus
Move the AX cursor without clicking.
→ verb 09
ask_human
Pause and surface a question to you.
→ verb 10
done
Declare completion. Hand control back.
§ 03 · Built for power users

Six things you'll actually use.

A short list, on purpose. Browsie isn't trying to be a platform. It's a focused cockpit for the 80% of browser tasks that are tedious, repetitive, and obviously a job for software.

§
[ PLAN ]

Visible plan.

The subgoal checklist is right there in the side panel. As the agent works, items tick green. When it's stuck, it tells you which one — not just "I failed."

[ STREAM ]

Live step log.

Every action with the agent's one-line thought, color-coded by outcome. Scroll back through what it did, why, and what came back from the page.

[ TASKS ]

Saved tasks.

Promote any goal to a reusable task. Schedule it to run on an interval or at a future time. Re-run by name. The morning routine writes itself.

[ EXTRACT ]

Typed extraction.

Ask for one field, get a string. Ask for a list, get a CSV. The accessibility-tree snapshot keeps the agent honest about what's actually on the page.

[ LIBRARY ]

Your library.

Name, email, address, freeform preferences — saved locally, sent to the LLM only when the form needs them. The agent fills with your real data, not placeholders.

[ STOP ]

Hard stops.

Refuses to submit, send, or buy by default. Click-loop, no-effect, and step-cap detectors auto-pause when the agent is spinning. You stay in the loop.

§ 04 · Privacy posture

Your machine. Your terms.

A browser agent reads everything on the page. That's a lot of trust. Browsie's answer is to never have the data in the first place — no cloud, no telemetry, no analytics, no contact with anyone but the LLM endpoint you point it at.

  • 01 / 04

    Local LLM by default.

    Configured for LMStudio / Ollama / any OpenAI-compatible endpoint. The default points at localhost. Page snapshots travel exactly as far as your inference server.

  • 02 / 04

    Storage stays local.

    Your library, settings, saved tasks, and run history live in chrome.storage.local. Never synced. Never uploaded.

  • 03 / 04

    No telemetry. No analytics.

    Zero outbound calls except (a) the LLM endpoint you configured and (b) the page the agent is currently working on. There are no third parties.

  • 04 / 04

    You can read every line.

    Browsie is unpacked-loadable in dev mode. Open the folder. Read the source. The whole agent fits in one short evening.

§ 05 · Get it running

Five minutes,
unpacked.

Browsie is distributed as an unpacked Chrome / Edge extension. Drop the folder in, point it at your local model, type a goal, and watch.

step 01
Unzip.
Drop the folder anywhere on your machine. Don't move it after — the extension references it by path.
step 02
Load unpacked.
Open chrome://extensions. Enable Developer mode. Click Load unpacked and select the folder.
step 03
Point at your LLM.
Click the Browsie icon, open [CFG], paste your endpoint and model name. LMStudio default works out of the box.
step 04
Type a goal.
"Fill the contact form on this page with my work email." Click Execute. Watch the plan tick.