Skyvern is not trying to be a generic agent wrapper. Its homepage says the quiet part out loud: “AI agents to automate workflows on any website,” with support for CAPTCHAs, 2FA, proxies, data extraction, document processing, form filling, and testing. That is a very specific bet, browsers are the integration layer when APIs are missing, bad, or slow.
The shipping signal is the louder story. ToolVitals shows 20 release events in 30 days and 21 releases in 90 days, on a repo with 21,493 GitHub stars. The latest release, v1.0.32, reads like a hardening pass, not a vanity bump. It includes fixes around reviewer LLM calls, copilot error handling, run detail page access after workflow deletion, and docs cleanup around SQLite-first self-hosting and env-file handling.
That combination matters. Skyvern is spending code on reliability, self-host clarity, and workflow continuity, which is exactly where browser automation tools usually fall apart in production. The homepage also pushes developer ergonomics hard: Python and TypeScript SDKs, Docker self-hosting, MCP support, webhooks, and an explainability layer that summarizes what the AI did and why.
What Skyvern says it is
The current positioning is consistent across the site and the repo. Skyvern says it is open source, it says it can run on-premise, and it says it can automate real browser workflows without custom scripts or APIs. The site also claims 30,000+ users and customers, but that is a marketing claim, not something ToolVitals can independently verify from code or releases.
The release page reinforces the same direction. v1.0.32 is small on drama and big on operational polish. The notes point to guardrails for copilot flows, reviewer logic, and self-host documentation. That is what mature automation software looks like when the team expects people to run it against messy portals, not toy demos.
How to read the numbers
Skyvern’s 20 release events in 30 days are strong, but not the highest in the set. n8n has 51 release events in 30 days, and OpenClaw has 48. Skyvern is still moving fast, just from a smaller base, with 21,493 stars versus n8n’s 186,633 and OpenClaw’s 368,112. In other words, Skyvern is not the biggest name in the room, but it is shipping like a team that expects the product to live in production.
What the data does not tell you
ToolVitals can see commits, releases, stars, uptime, and site claims. It cannot see whether Skyvern’s workflows actually succeed on ugly enterprise portals, whether the explainability layer is useful in practice, or whether users stick around after the first pilot. It also cannot tell you code quality, security posture beyond what is visible publicly, revenue, or customer satisfaction.
So the right reading is narrow and honest. Skyvern looks well maintained, aggressively shipped, and clearly focused on browser-based automation for sites without APIs. That is enough to say the product is real. It is not enough to say it is the best choice for every automation stack.
Bottom line
If your team spends time logging into vendor portals, scraping data behind auth walls, or automating forms that break when the DOM shifts, Skyvern is worth a serious look. The release cadence and homepage positioning both suggest a team building for production browser work, not just agent theater. If your workflows already have clean APIs, Skyvern is probably the wrong hammer.