Understand the page
Read the DOM and accessible controls instead of guessing from screenshots.
OPEN-SOURCE · PLAYWRIGHT · LOCAL-FIRST
Your browser.
Your model.
Your call.
Give your browser a goal. FreeComputerUse lets an AI plan the steps while Playwright runs them on your machine, with website access and sensitive actions kept in your hands.
Bring an API key or sign in with a supported model CLI.
FREECOMPUTERUSE / LOCAL RUNON YOUR MACHINE
GOALCompare three options and save the evidence
- 01ObserveRead page structure and controls✓
- 02PlanChoose a small, checkable action batch✓
- 03ExecuteUse the approved browser tab•••
- 04VerifyCheck results before you reuse them
YOU STAYIN CONTROL
01Runs locallyPlaywright drives your browser
02Works across modelsChoose an API or supported CLI
03Asks before direct navigationApprove a new origin before it opens
A SMALL, CHECKABLE LOOP
You set the goal.
The agent handles the clicks.
The model suggests browser actions. Your local app executes them, checks what changed and shows you the run.
Keep each batch small
Ask your chosen model for the next bounded set of browser actions.
Act in your browser
Playwright runs locally. Direct navigation to a new origin asks for approval.
Check before reuse
Review the result, repair failures and replay compatible learned workflows later.
Learned repeats can run with zero model calls when the workflow still matches.
ONE RECORDED RUN · NOT A GENERAL SUCCESS CLAIM
See what the steps
look like in practice.
A local browser run investigates a synthetic API incident, checks six pages and saves a report.
204
RECORDED PLAYWRIGHT SESSIONAPPROVAL SHOWN
21 successful actions · 10 model calls · $0.00650 configured estimateOpen full recording ↗
PICK THE MODEL THAT FITS YOU
One browser agent.
More than one way to connect.
Use provider API keys for flexible model access, or authenticate with a supported local CLI subscription. The available choices depend on your configuration.
01 API KEY
Connect a model API
Set the provider endpoint, model and key in your local environment.
OpenAI-compatible endpoints work through a configurable base URL. Anthropic has its own API mode.02 SUPPORTED CLI SIGN-IN
Use a subscription
Sign in locally with the provider CLI, then select its subscription mode.
CLI sign-in uses that provider’s own plan limits. API billing and subscription usage are separate.Provider setup and evidence: model guide · support matrix. Some provider routes have not been live-tested in this project.
START WITH A REAL GOAL
A task library for
work worth automating.
Copy a goal into your local workspace. Browse read-only research tasks or switch to synthetic workflows for practice. Source checks and agent trials are labelled separately.
Explore focused browser checks
The original small pages remain available for checking individual interactions.
LOCAL CONTROL, MADE VISIBLE
The browser stays yours.
Browser execution, profiles, history and downloads stay on your machine. The goal and page context needed to plan a task go to the model provider you chose.
- Approve each website origin
- Confirm sensitive actions separately
- Inspect action history, checks and repairs
- Replay compatible workflows without model calls
Local storage is not encrypted. Immediate cross-origin redirects from newly opened popups remain a known limitation. Read the security boundaries before using personal or sensitive data.
GET RUNNING LOCALLY
Bring your own model.
Keep your browser close.
Requires Node.js 22.13+, npm and a browser Playwright can launch. An installed Chrome or Edge can be selected as a best-effort option without a Playwright browser download.
Open setup instructions01 / TERMINALgit clone https://github.com/OthmaneBlial/FreeComputerUse.git
cd FreeComputerUse
npm ci
cp .env.example .env
chmod 600 .env
FCU_BROWSER_CHANNEL=chrome npm run devThen open http://127.0.0.1:4318