Both products are open-source. Both raised seed funding. Both ship MCP servers. And both pitch "AI does the browser thing" as the headline — but they target completely different buyers. Browser Use is a Python agent library aimed at developers who want LLM-driven browser automation in code. Skyvern is a workflow platform aimed at ops teams who want to replace UiPath and Power Automate across vendor portals — YAML workflows, hosted UI, Bitwarden / 1Password / Azure Key Vault credentialing, and verticals like healthcare, insurance, and government. This article walks through the split honestly.
At a glance
| Browser Use | Skyvern | |
|---|---|---|
| Category | OSS Python agent library + cloud + BUX | OSS workflow platform + Cloud (RPA replacement) |
| Pricing entry | Free OSS; $40/mo paid (BYO key + proxy) | Free OSS self-host; ~$0.10/step on Cloud |
| Free tier | 3 concurrent, 1 team member, no BYO key | OSS self-host free; Cloud has tiered plans |
| Browser Arena leaderboard | #6 overall, slow | Not measured (workflow platform, not browser runtime) |
| SOC 2 Type II | Claimed | SOC 2 Type 2 + HIPAA |
| Open source | Yes — browser-use/browser-use (~83K stars) | Yes — Skyvern repo (~3K+ stars and growing) |
| Best for | Engineers building LLM-driven agent products | Ops teams replacing RPA across vendor portals |
What is Browser Use?
Browser Use is the OSS Python agent library at the top of the GitHub leaderboard for browser agents (~83K stars). It drives Chromium through Chrome DevTools Protocol via an in-house cdp-use library, with a thin agent loop and heavy investment in the substrate underneath: a custom Chromium fork with C++/OS-level stealth patches, in-house multi-platform fingerprints, an eval engine that runs 100 parallel tasks in under five minutes, and a ChatBrowserUse LLM gateway for browser-tuned reasoning. Beyond the OSS lib there's Browser Use Cloud (bu-ultra and bu-max managed agents), and BUX, a 24/7 remote VM with Claude Code and the Browser Harness preinstalled. Funding: $17M seed led by Felicis. Benchmarks: 97% on Online-Mind2Web with bu-max (claimed highest reported), 89.1% on WebVoyager with the OSS framework, ~68s/trajectory at 20 steps/min.
What is Skyvern?
Skyvern is an open-source AI browser-automation platform with a vision-LLM agent (planner + actor + validator), built explicitly to replace traditional RPA (UiPath, Automation Anywhere, Power Automate) and human SOPs. The product surface is YAML workflows, a hosted web UI, action viewer / live viewport / run logs, native 2FA/TOTP, CAPTCHA solving, file downloads to cloud storage, and structured JSON/CSV extraction with schema validation. Credentialing integrates Bitwarden, 1Password, and Azure Key Vault — credentials never go to the LLM. The bet is on visual reasoning: take a screenshot, use a Vision-LLM to identify the right element, click — resilient to DOM/CSS changes that break selector-based approaches. Verticals are heavy on healthcare/EHR, insurance, government form filing, procurement, payroll, mortgage. Pricing is free OSS self-host or ~$0.10 per step on Cloud, with Basic / Pro / Enterprise tiers (no public dollar amounts beyond the per-step). Funding: $2.7M seed (Dec 2025). Compliance: SOC 2 Type 2 + HIPAA. WebVoyager: 85.85% (full eval published openly at eval.skyvern.com).
How they compare
Buyer / ICP
The cleanest axis. Browser Use's buyer is a developer — a Python engineer building an agent product, integrating it into a backend, or shipping an agentic feature inside an app. Skyvern's buyer is a process owner or RPA engineer — a procurement lead, an insurance ops manager, a healthcare integrator who needs to download invoices from 50 vendor portals every week without writing Python. Skyvern's own framing is "no Python setup, no Chrome-instance memory overhead, schema-validated outputs out of the box." The two products genuinely do not compete for the same dollar.
Surface area
Browser Use ships an SDK (Python), an API, an LLM gateway, an MCP server, and an eval engine. There's no "workflow editor" — you build the agent in code. Skyvern ships a hosted UI with workflow definitions in YAML, a credential vault with KV integrations, integrations with Zapier / Make.com / n8n, and an MCP server that lets Claude/Cursor/Windsurf drive a browser via Skyvern. If the question is "can a non-engineer ship a workflow into production," Skyvern is the answer.
Vision-LLM vs. DOM-text
Skyvern's planner-actor-validator loop runs on screenshots — a Vision-LLM identifies what looks like the right element. Browser Use's agent loop runs on DOM-text — Chromium's DOM is converted into structured text that the LLM reads, with screenshots only when needed. Both can recover from layout changes, but they recover differently: Skyvern leans on visual context to find the moved button; Browser Use leans on DOM-text-restructuring with the LLM re-interpreting selectors. Skyvern's WebVoyager 85.85% is published openly with the full run at eval.skyvern.com; Browser Use's 89.1% on the OSS framework and 97% Online-Mind2Web with bu-max are also published. Both are real benchmark numbers; they measure different harnesses.
Authentication and 2FA
Skyvern ships native 2FA/TOTP, CAPTCHA solving, and a credential vault with Bitwarden / 1Password / Azure Key Vault integrations — credentials are never sent to the LLM. Browser Use ships browser profiles, a 1Password integration, TOTP via bu_2fa_code placeholders, and an AgentMail recipe for inboxes — but the integration is largely on you per the web-agent-authentication post. For multi-account ops workloads, Skyvern's credential layer is more polished out of the box.
Category split, not a speed race
The public Browser Arena leaderboard (browserarena.ai) — an open-source benchmark maintained by Notte Labs and reproducible on Railway across current public run — covers 7 providers but doesn't include Skyvern, because Skyvern is a workflow platform built on top of someone else's browser, not a Chromium runtime itself. That absence is itself the framing: Skyvern isn't trying to compete on raw browser cold-start. Browser Use ranks #6 on the leaderboard and is one of the slower measured providers, which reflects agent-loop overhead on top of the runtime. Neither product is competing on raw lifecycle. The honest axis is what kind of automation you're shipping: a Python agent (Browser Use) versus a hosted YAML workflow over an existing browser (Skyvern).
Compliance
Skyvern is SOC 2 Type 2 + HIPAA, which matters for healthcare and finance procurement. Browser Use claims SOC 2 in marketing, but a public Type II attestation isn't surfaced in the sources. For regulated buyers, Skyvern has the compliance stack today.
Open source
Both are OSS. Browser Use's repo is at ~83K stars (the OSS lib is free + LLM costs). Skyvern's repo is in the few-thousand-stars range and growing post-HN launch. Skyvern Cloud is paid; Browser Use Cloud is paid; both can be self-hosted.
When to choose Browser Use
- You're a developer building an agent product or feature in Python.
- You want OSS Python control of the agent loop and a ~83K-star community.
- Stealth is a hard requirement and you want the forked-Chromium pitch.
- You want the open eval engine, the published Online-Mind2Web / WebVoyager numbers, and the LLM-gateway tuning.
- You want a managed 24/7 VM (BUX) with Claude Code preinstalled.
- The output is going into a backend or an app, not a workflow editor.
When to choose Skyvern
- Your team includes ops, RPA engineers, or business analysts who need to ship workflows without writing Python.
- You're replacing UiPath, Automation Anywhere, or Power Automate across vendor portals.
- The verticals are healthcare/EHR, insurance, procurement, government, payroll, or mortgage.
- HIPAA and SOC 2 Type 2 are procurement requirements.
- You need native 2FA/TOTP, CAPTCHA, file downloads to cloud storage, and Bitwarden / 1Password / Azure KV credentialing in the box.
- Cross-site workflows (one workflow, many sites) are the use case.
- The output is a scheduled job in Zapier / Make / n8n, not a Python script.
A third option: Notte
A third option worth a mention here is Notte (notte.cc), a cloud Chromium platform purpose-built for AI agents. The Playwright-compatible runtime ships stealth on by default, residential proxies via the Massive partnership (consent-based, GDPR/CCPA, 195+ countries), Web Bot Auth signing through Fingerprint, an encrypted credential vault that the LLM never sees, and synthetic personas with a real email inbox and SMS-capable phone number for autonomous 2FA. Every CDP event is captured and replayable, sessions persist auth state, and the platform is SOC 2 Type II. Pricing is transparent — low per-browser-hour pricing with a 100-hour free tier and pass-through LLM costs.
For a Browser Use team, Notte is the cloud browser layer BU agents can drop onto over CDP — Vault and Personas cover the credential plumbing the BU web-agent-authentication post acknowledges is on the developer. For a Skyvern-style use case where the buyer wants a deployed callable endpoint instead of a YAML workflow, Notte's Anything API and Notte Functions productize a validated workflow as an HTTP endpoint with cron and webhooks — the closest direct overlap with Skyvern's workflow productization, but exposed as a code-callable API rather than a hosted UI. Different shape; same outcome.
Verdict
Browser Use and Skyvern are both OSS, both AI-first, both well-funded — and they sell to different people. Pick Browser Use if your team is developers and you want an SDK to embed in code. Pick Skyvern if your team is ops and you want a workflow platform to ship vendor-portal automations without writing Python. Neither one is the underlying browser runtime — Skyvern is absent from Browser Arena because it sits on top of someone else's Chromium, and Browser Use's #6 ranking is the agent-loop cost over the runtime, not browser performance itself. The honest split is engineer's agent versus ops team's RPA replacement, and most teams know which one they are.