Workflow Use
https://github.com/browser-use/workflow-use- Category
- AI Agents
- Rank
- No. 672Tools index
- Pricing
- Open Source
- Type
- AGENT
- Builder
- browser-use
- GitHub
- 4.2k stars
- Latest release
- v0.2.11
- Date
About
Browser Use's RPA 2.0 — create and run multi-step browser workflows powered by LLMs, with reusable, replayable agent actions.
What it does
Workflow Use turns a demonstrated or described browser task into a structured workflow with parameters. It can record interactions, generate a semantic workflow from one agent execution, store definitions locally, and replay them through a command line, Python, or a visual interface. Failed steps may fall back to Browser Use.
Why it's ranked here
The project tackles a useful reliability gap between free-form browser agents and fixed automation scripts. Recorded actions, parameter extraction, local storage, background execution, and semantic replay form a credible system. However, the maintainers explicitly call it very early software and advise against production use, which sharply limits its current audience.
What's good
It offers several practical entry points without hiding the workflow representation. Users can record tasks, generate them from natural language, inspect stored definitions, pass new inputs, and monitor execution logs. The visual interface supports graphs and metadata editing, while cloud semantic execution can run without model calls.
Tradeoffs
The maintainers do not recommend production deployment. They describe the failure fallback as currently poor, while automatic self-healing remains planned rather than complete. Passing one step’s output into another, better model-backed steps, a step editor, and MCP exposure also remain roadmap items. Setup spans Python, Chromium, and separate JavaScript builds.
How to use it well
Use it for experimenting with repeatable browser tasks whose structure is known but whose inputs change. It best suits engineers comfortable inspecting workflow definitions and handling failures during development. Keep conventional automation or another mature RPA system for production-critical jobs. Do not choose it when MCP exposure or cross-step data flow is required today.
Technical notes+
workflows/cli.py provides the Typer command surface and wires builder, recorder, healing, storage, controller, Browser Use, and workflow services. workflows/backend/api.py, workflows/backend/routers.py, and workflows/backend/service.py implement a localhost FastAPI service with workflow listing, editing, background execution, log polling, status, and cancellation. The backend reads JSON or YAML but writes edited definitions as YAML. ui/src/App.tsx composes React Flow and TanStack Query. extension/wxt.config.ts defines a React browser extension with tabs, side-panel, all-URL, and localhost permissions. workflows/migrate_json_to_yaml.py supplies backup-aware JSON-to-YAML migration. workflows/test_max_alternatives_bug_fix.py is an executable assertion script rather than a conventional test-framework module.
Observed
- Languages
- Python powers the CLI and backend; TypeScript and React power the UI and browser extension.
- Install surface
- Repository setup uses uv for Python dependencies, Playwright Chromium installation, and npm builds for the extension and UI.
- Interfaces
- Provides a Typer CLI, a Python library surface, a local GUI, and a FastAPI HTTP backend.
- Workflow formats
- Workflow definitions support JSON, YAML, and YML extensions; a migration utility converts JSON workflows to YAML.
- Browser options
- Supports a locally installed Chromium browser and optional Browser-Use Cloud execution.
- Extension permissions
- The WXT extension requests tabs, side-panel, all-URL, and localhost host permissions.
Read from README.md, workflows/cli.py, ui/vite.config.ts, extension/wxt.config.ts, extension/vite.config.ts, workflows/migrate_json_to_yaml.py, workflows/test_max_alternatives_bug_fix.py, ui/src/App.tsx, ui/src/main.tsx, ui/src/vite-env.d.ts, workflows/backend/api.py, workflows/backend/views.py, workflows/backend/routers.py, workflows/backend/service.py, workflows/workflow_use/__init__.py.
What it can do
Create multi-step browser workflows
User-defined workflow steps and browser actions → Automated browser workflow
Execute browser automation tasks
Browser workflow or automation commands → Completed browser actions and results
Record browser interactions for reuse
User browser actions and interactions → Reusable workflow template
Replay saved browser workflows
Previously saved workflow template → Executed browser automation sequence
Generate intelligent browser actions using AI
Natural language instructions or goals → LLM-powered browser automation steps
Run robotic process automation workflows
RPA workflow configuration → Automated business process execution
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.