Scrapybara OSS

Note

This is a simplified extraction of our internal backend. Many features were removed and there may be rough edges or bugs. Please report issues via GitHub Issues.

This folder contains two minimal, decoupled open-source components:

• act/ — /v1/act endpoint compatible with existing Scrapybara SDKs
• computer/ — xdotool mouse/keyboard/screenshot API

This repo is intentionally minimal. If you are building a production system, consider extracting the core logic and rebuilding APIs around it for your infra:

• LLM action space core: act/src/providers.py
• Computer-use core: computer/src/computer.py
• Image/runtime: computer/docker/ (Dockerfile + entrypoint)

Getting started

1. Start the Computer service

cd computer
docker build -t scrapybara-computer -f docker/Dockerfile .
docker run --rm \
  -e WIDTH=1024 -e HEIGHT=768 -e DISPLAY_NUM=1 \
  -p 8000:8000 -p 5900:5900 -p 6091:6091 \
  scrapybara-computer

Stream is available at http://localhost:6091

2. Start the LLM service

uv sync
export OPENAI_API_KEY=...
export ANTHROPIC_API_KEY=...
uv run uvicorn act.src.main:app --host 0.0.0.0 --port 8001 --reload

3. Expose the API with ngrok and run the SDK example

ngrok http 8001
export SCRAPYBARA_BASE_URL="https://*.ngrok-free.app"
export SCRAPYBARA_API_KEY=...
uv run python main.py

License

Apache-2.0