// Agent profile
Browserbase Skills is the open-source skills bundle that pairs Browserbase's cloud browser runtime with Claude Code's harness. Installable via `/plugin install browse@browserbase` from the Claude Code marketplace, the bundle exposes nine specialized skills covering the full browser-automation surface: web interaction (browser), platform CLI (browserbase-cli), serverless deployment (functions), automation diagnostics (site-debugger), DevTools capture (browser-trace), usage forecasting (bb-usage), authenticated cookie sync (cookie-sync), static-page fetch (fetch), and structured web search (search). The strategic positioning is precise: Browserbase did not try to be a coding-agent runtime; it built tools that the agents *already in production* (Claude Code, Cursor SDK, OpenAI Agents SDK) can call. Open source on top, paid runtime underneath — every skill is publicly readable, but the cloud sessions, residential proxies, anti-bot fingerprints, and Functions runtime are Browserbase's commercial moat. As of May 2026, the GitHub repo holds 1,557 stars and is anchor of the browser-tool layer in the agent-substrate stack.
Browserbase Skills is the substrate that picked the smallest fight in the harness-substrate landscape and is winning it cleanly. Where Cursor SDK competes for the engineering-pipeline runtime slot and OpenAI’s Apps SDK competes for the consumer-distribution slot, Browserbase competes only for the browser-tool layer — the layer every agent that needs to log in, click around, and extract data from web sessions has to traverse.
The strategic move is to deploy as a Claude Code plugin, not as a competing harness. Every Claude Code user becomes a potential customer without Browserbase having to win them away from Anthropic’s distribution. The companion product, Stagehand, is the open-source SDK for browser agents (22.4K stars). The business model is the cloud runtime underneath: scale-out browser sessions, residential proxies, function-style deployments, anti-bot evasion.
| Skill | Purpose |
|---|---|
browser | Web interaction via remote Browserbase sessions with anti-bot stealth |
browserbase-cli | Direct platform API — sessions, projects, contexts, functions |
functions | Serverless browser deployment to Browserbase cloud |
site-debugger | Diagnostics for failing automations — bot detection, selectors, timing |
browser-trace | DevTools protocol capture with searchable per-page buckets |
bb-usage | Terminal dashboard for usage stats and cost forecasts |
cookie-sync | Chrome-to-context cookie sync for authenticated access |
fetch | Static page retrieval without browser sessions |
search | Web search returning structured results |
Freemium. The skills bundle and the Stagehand SDK are free and open source. The paid layer is the Browserbase platform: per-session and per-page pricing for cloud browser sessions, with separate tiers for residential proxies, captcha resolution, and Functions deployments. The bb-usage skill surfaces forecasted spend in the terminal — useful for scoping CI usage before it goes to production.
The lowest of the three major harness substrates. Stagehand is Apache-licensed; you can self-host browser sessions; you can swap the Claude harness underneath the skills. The lock-in lives entirely in Browserbase’s paid runtime — anti-bot evasion, fingerprint randomization, residential-proxy pools — which is genuinely difficult to replicate in-house.
If your agents only need to read public pages, you can leave anytime. If they need to log into customer SaaS at scale or evade bot detection on real-world e-commerce sites, Browserbase’s runtime is the default.
Browserbase Skills is best understood not as a competitor to Cursor SDK or OpenAI Apps SDK but as a complement to both. The most likely production agent stack in late 2026 uses Cursor SDK or OpenAI Agents SDK as the orchestration layer and Browserbase Skills as the browser-tool layer when the agent needs to operate web sessions. See our comparison of the three harness substrates for the positioning matrix.
Open protocol for connecting AI agents to frontend applications — a universal contract for streaming events, tool calls, and state between agents and UIs.
Rust-based browser automation CLI from Vercel Labs that lets AI agents drive browsers one command at a time via accessibility-tree snapshots.
Rust-based desktop computer-use framework from Vercel Labs — agents control apps via accessibility trees instead of screenshots.