Skyvern
Open-source browser automation framework using computer vision and LLMs to navigate websites like humans, eliminating fragile selectors.
6,500+
GitHub Stars
30,000+
Users
2023
Founded
Overview
Skyvern automates browser workflows by using computer vision and LLMs to interact with websites the way humans do—no brittle XPath selectors or manual maintenance. Teams use it to automate repetitive portal tasks: downloading invoices, filling forms, copying data, and navigating multi-step workflows. Available as Python/TypeScript SDKs or self-hosted via Docker, Skyvern integrates with agent frameworks through Model Context Protocol (MCP) and provides webhooks for workflow orchestration.
The Verdict
Who Should Use Skyvern?
Best For
- Teams automating repetitive web portal tasks across multiple sites
- Organizations eliminating manual data entry and form filling
- Companies frustrated with fragile selector-based automation
- Developers building AI agents that need browser interaction capabilities
- Businesses wanting self-hosted automation infrastructure
Not Ideal For
- High-speed scraping requiring sub-second response times
- Simple static page scraping without dynamic interaction
- Teams seeking fully managed cloud-only solutions
What's Great
- Vision + LLM approach works across websites without site-specific code
- Self-hostable via Docker—full control over data and infrastructure
- Python and TypeScript SDKs integrate with existing codebases
- Model Context Protocol (MCP) ready for Claude, GPT, and Gemini agents
- Webhooks and event streaming for workflow orchestration
- Open-source community with active development and commercial support option
Watch Out For
- LLM-based automation slower than traditional selector-based scripts
- Costs associated with LLM API usage (OpenAI, Anthropic) when self-hosting
- Computer vision may struggle with highly dynamic or poorly designed UIs
- Requires infrastructure management expertise for production deployment
Pricing
Open Source
Free
Self-host with Docker Compose. Bring your own LLM API keys. Full feature access with no restrictions or usage limits.
Cloud (Beta)
Contact
Managed hosting with built-in LLM access, infrastructure management, and support. Currently in beta—contact for pricing.
Enterprise
Custom
Dedicated infrastructure, custom SLAs, white-glove support, compliance certifications, and professional services.
View all features & details
Key Features
- Computer vision + LLM browser automation
- Python SDK: pip install skyvern
- TypeScript SDK with type safety
- Self-hostable via Docker Compose
- Model Context Protocol (MCP) support
- Webhooks and event streaming
- REST API for programmatic control
Platforms
- Python 3.10+
- Node.js/TypeScript
- Docker containers
- Self-hosted deployment
- Cloud (managed) - beta
How It Compares
| Feature | Skyvern | Playwright MCP | Browserbase |
|---|---|---|---|
| Automation Method | Vision + LLM hybrid | Accessibility tree | Traditional + stealth |
| Selector Brittleness | Low (visual understanding) | Low (semantic structure) | High (requires selectors) |
| Deployment | Self-hosted + cloud | Self-hosted only | Cloud-only |
| Speed | Slower (LLM inference) | Fast (structured data) | Fast (direct control) |
| Best For | Complex multi-site workflows | MCP agent integration | Managed browser infrastructure |
User Reviews
Loading reviews...