Browser Use on Render

Deploy a simple Python API that runs Browser Use tasks on demand.

What you get

• A FastAPI service with POST /run to execute a task and GET /health for checks.
• A render.yaml blueprint ready for one-click deploys.

Deploy to Render

1. Fork this repo.
2. In the Render dashboard, choose Blueprint and select your fork.
3. Add BROWSER_USE_API_KEY during setup.
4. Deploy.

Environment variables

LLM_PROVIDER: LLM provider to use. Defaults to anthropic. Supported values: browser_use, openai, anthropic, google. OpenAI is recommended for best reliability on Render.

LLM_MODEL: Model name for the selected provider. Required for most providers.

BROWSER_USE_API_KEY: Your Browser Use Cloud API key. Required when LLM_PROVIDER=browser_use. Browser Use quickstart

Other provider keys are optional and only required when selected. Supported values and required vars are listed in the Browser Use docs. Supported models

Using a local .env file

Copy the example and fill in the values:

cp env.example .env

Then edit .env with your preferred provider:

# Recommended: OpenAI (most reliable on Render)
LLM_PROVIDER=openai
LLM_MODEL=gpt-4o-mini
OPENAI_API_KEY=your_key_here

# Alternative: Anthropic
# LLM_PROVIDER=anthropic
# LLM_MODEL=claude-3-5-sonnet-latest
# ANTHROPIC_API_KEY=your_key_here

# Alternative: Google
# LLM_PROVIDER=google
# LLM_MODEL=gemini-2.0-flash
# GOOGLE_API_KEY=your_key_here

API usage

Run a task:

curl -X POST "$SERVICE_URL/run" \
  -H "Content-Type: application/json" \
  -d '{"task":"Find the number 1 post on Show HN"}'

Get the full agent result:

curl -X POST "$SERVICE_URL/run" \
  -H "Content-Type: application/json" \
  -d '{"task":"Find the number 1 post on Show HN","response_mode":"full"}'

Health check:

curl "$SERVICE_URL/health"

Example queries

Get the top trending repo on GitHub:

curl -X POST "$SERVICE_URL/run" \
  -H "Content-Type: application/json" \
  -d '{"task":"Go to github.com/trending and tell me the #1 trending repo"}'

Check the top story on Hacker News:

curl -X POST "$SERVICE_URL/run" \
  -H "Content-Type: application/json" \
  -d '{"task":"Go to news.ycombinator.com and tell me the title of the #1 story"}'

Search and summarize:

curl -X POST "$SERVICE_URL/run" \
  -H "Content-Type: application/json" \
  -d '{"task":"Search Google for the latest SpaceX launch and summarize what you find"}'

Fill out a form:

curl -X POST "$SERVICE_URL/run" \
  -H "Content-Type: application/json" \
  -d '{"task":"Go to httpbin.org/forms/post and fill out the form with fake data, then submit it"}'

Example use cases

• Collect a daily summary of a page that changes often.
• Find and extract the top item from a list or table.
• Verify a UI flow and report any errors.
• Log in to a site and capture a specific value for a report.

Local development

1. Install dependencies:

pip install -r requirements.txt

2. Set your key:

export BROWSER_USE_API_KEY=...

3. Run the server:

uvicorn app.main:app --host 0.0.0.0 --port 8000

Docker (optional)

This repo includes a Dockerfile based on the Playwright Python image so Chromium and dependencies are already available.

Render plan recommendations

This template runs Chromium inside the container, which requires memory. Here's what to expect:

Plan	RAM	Best for
Free	512 MB	Simple sites (GitHub, Hacker News, static pages)
Starter	512 MB	Same as Free, with more CPU for faster processing
Standard	2 GB	Heavy JavaScript sites (weather.com, complex SPAs)

Note: Sites with heavy JavaScript, anti-bot measures, or lots of ads may show empty pages or time out on the Free and Starter tiers. If you encounter this, try the Standard plan.

LLM provider notes

• OpenAI is the most reliable provider from Render's network.
• Anthropic may experience intermittent connection timeouts on lower-tier plans.
• Browser Use Cloud offloads browser execution to their servers, reducing memory usage on your Render instance.

Notes

• This template uses the Python example from the Browser Use quickstart. Browser Use quickstart