- Agent Chat + Multimodal - Text/image chat with streaming responses.
- Tool-Driven Agent Loop - Built-in tools for context, diary, profile, tracking, device, web search, screen read, and browser actions.
- Computer Use Baseline - Manual screenshot input + autonomous
screen_get; browser state read and controlled/external navigation. - Long-Term Memory - OpenViking-compatible file-memory structure.
- Continuous Tracking - Target tracking, snapshots, and change records.
- Desktop Runtime - Electron shell for local agent usage.
backend/- FastAPI API, SQLAlchemy models, services, schedulers.backend/tests/- Pytest suite.frontend/- React 19 + Vite + TypeScript app.desktop/- Electron runtime and packaging.docs/- Architecture, plans, and test notes.data/- Local runtime data and file-memory workspace.
cd backend
python -m pip install -r requirements.txt
python -m uvicorn app.main:app --reload --port 8000cd frontend
npm install
npm run devcd desktop
npm install
npm run dev# Backend tests
cd backend
pytest -q
# Frontend build
cd frontend
npm run build
# Desktop package
cd desktop
npm run dist- Core security/config:
MERCURYDESK_SECRET_KEYMERCURYDESK_FERNET_KEYMERCURYDESK_DATABASE_URLMERCURYDESK_CORS_ORIGINS
- Agent loop:
MERCURYDESK_AELIN_AGENT_LOOP_ENABLEDMERCURYDESK_AELIN_AGENT_LOOP_MAX_ROUNDS
- Browser tools:
MERCURYDESK_BROWSER_TOOL_CDP_ENABLEDMERCURYDESK_BROWSER_TOOL_CDP_ENDPOINTMERCURYDESK_BROWSER_TOOL_OPEN_EXTERNAL_ON_NAVIGATE
Do not commit API keys, OAuth secrets, or local database artifacts.
- Chinese README: README.zh-CN.md
- English README: README.en.md
- Contributor guide: AGENTS.md
- Backend notes: backend/README.md
- Feishu remote control: docs/feishu_remote_control_v1.md
- Docs index: docs/INDEX.md
- Manual test cases: docs/agent_loop_manual_test_cases.md