Browser Use
Open-source library that lets AI agents control a real browser to automate and extract from any site
Open-source AI agent that automates browser workflows with vision instead of brittle selectors
Skyvern is a Y Combinator-backed, open-source (AGPL-3.0) AI agent that automates browser workflows from natural language, using vision language models instead of brittle DOM selectors so it survives layout changes. It handles logins, 2FA, and CAPTCHAs, and extracts data to CSV or JSON. With 22,000+ GitHub stars, it offers a free cloud tier plus Hobby at $29/mo and Pro at $149/mo; the open-source version is free to self-host. Best for teams automating web tasks that break traditional scrapers.
Skyvern is an open-source AI agent that automates browser-based workflows from a natural-language description. Instead of the DOM, XPath, or CSS selectors that traditional automation and scraping tools rely on — and which break the moment a site tweaks its layout — Skyvern uses vision language models and computer vision to read a page the way a person would. That makes its workflows self-healing: they keep working through UI changes without constant maintenance. You can describe a task, upload a standard operating procedure, or use a record-once mode to demonstrate a flow and have Skyvern repeat it.
Beyond navigation, Skyvern handles the parts of the web that stop simpler bots: full login flows with 2FA and TOTP, a stored-credential vault, native CAPTCHA solving, and residential or datacenter proxies with geo-targeting. It extracts data into any schema as CSV or JSON, and includes OCR and document classification. The project is Y Combinator-backed, licensed AGPL-3.0, and has passed 22,000 GitHub stars; it ships as a cloud service, a REST API with Python and TypeScript SDKs, and a self-hostable Docker image that works with OpenAI, Anthropic, Gemini, or local models via Ollama.
Skyvern suits teams that need to automate real web tasks — the kind involving authentication, forms, and data extraction across sites that actively resist scraping. Its open-source core is free to self-host for technical teams comfortable with Docker and their own LLM keys, while the cloud tiers trade that setup for a managed service billed by per-action credits.
Starting price: $29/mo · Free tier: yes · Model: freemium
Price history tracked from June 2026
| Plan | Price | Includes |
|---|---|---|
| Free | Free | Free monthly cloud credits, no credit card · 1 concurrent run · Datacenter proxies, webhooks, community support |
| Hobby | $29/mo | ~30,000 credits/mo (roughly 1,200 actions) · 10 concurrent runs · Basic CAPTCHA solver, stored credentials, email support |
| Pro | $149/mo | ~150,000 credits/mo (roughly 6,200 actions) · 25 concurrent runs · Advanced CAPTCHA solver, residential proxies, 2FA/TOTP, team workspaces |
| Enterprise | Custom | Unlimited credits and concurrent runs · Advanced anti-bot, SOC 2 Type II, HIPAA · Dedicated support |
| Open source | Free | AGPL-3.0 self-hosting via Docker · Bring your own LLM keys · Model-agnostic (OpenAI, Anthropic, Gemini, Ollama) |
| Pros | Cons |
|---|---|
| Open source (AGPL-3.0) with 22,000+ GitHub stars — free to self-host | Vision-LLM navigation can be non-deterministic and slower than scripted bots on complex sites |
| Vision-based navigation resists layout changes that break selector-based bots | Per-action credit costs add up on long workflows, pushing heavy use to the $149/mo Pro tier or Enterprise |
| Handles logins, 2FA, and CAPTCHAs that block simpler scraping tools | Self-hosting needs Docker plus your own LLM keys, and AGPL-3.0 may deter some commercial embedding |
| Flexible deployment: cloud, REST API, SDKs, and on-premises Docker |
Open-source library that lets AI agents control a real browser to automate and extract from any site
Autonomous general AI agent that plans and executes multi-step tasks end-to-end in the cloud
No-code AI agents that handle email, calendar, meetings, and 100+ app integrations
No-code platform for building and deploying autonomous AI agents for business workflows
Skyvern is open source under the AGPL-3.0 license, so you can self-host it for free with your own LLM keys. The hosted cloud service also has a free tier with monthly credits, then paid plans: Hobby at $29 per month and Pro at $149, plus custom Enterprise pricing.
Traditional RPA records brittle DOM or XPath selectors that break when a website's layout changes. Skyvern uses vision language models and computer vision to interpret the page like a person would, so workflows survive UI changes without constant selector maintenance. It also handles logins, 2FA, and CAPTCHAs natively.
Skyvern's cloud plans are Hobby at $29 per month (about 30,000 credits) and Pro at $149 per month (about 150,000 credits), with a free tier and custom Enterprise pricing above that. Credits are consumed per action, so cost depends on how many steps your workflows take. Self-hosting the open-source version is free.
Skyvern automates browser workflows described in natural language: logging in, filling forms, completing multi-step flows, handling 2FA and CAPTCHAs, downloading documents, and extracting data into CSV or JSON. It can be shown a task once via a record feature and then repeat it, and it connects through a REST API and SDKs.
Yes. Skyvern handles full authentication flows including 2FA and TOTP, offers a stored-credential vault, and includes CAPTCHA solving — a basic solver on Hobby and an advanced solver plus residential proxies on Pro. This lets it run automations on sites that block simpler scraping tools.
Vision-based navigation is more resilient to layout changes than hard-coded selectors, but it can be non-deterministic and slower than a scripted bot on multi-step or heavily defended sites. Per-action credit costs also add up on long workflows, so complex production use often lands on the Pro or Enterprise tier.