Browserbase Ecosystem
Browserbase is the cloud browser and saved browser-skill route when an agent needs repeatable website workflows, not just one-off local browsing or deterministic Playwright tests. Source: Ecosystem Capability Coverage, 2026-06-27; Skill Resolver, 2026-06-27
When To Use This
Use Browserbase when the workflow benefits from a managed browser session, a learned site procedure, or a reusable browser skill. Use Playwright first for CI regression tests and local deterministic assertions. Source: Skill Resolver, 2026-06-27
Decision Matrix
| Need | Route | Verification |
|---|---|---|
| Repeatable website workflow | Autobrowse (Browserbase Skills) / Browserbase skill family route | Saved skill plus replay smoke |
| Deterministic regression | Browser Testing Skills | Passing test |
| Visual/responsive QA | Browser Harness - Self-Healing Browser Automation or agent-browser | Screenshot and console check |
| Existing Chrome session | Chrome DevTools MCP | DOM/computed-style evidence |
Core Model
Browserbase belongs to browser automation. Localhost testing, browser QA, and agent search are separate capability concepts. Browserbase is strongest when discovery cost can be amortized into a skill. Source: User request, 2026-06-27; Browserbase Browse.sh (Open Web Catalog)
Browse.sh (Browserbase Skills) is the Browserbase skills catalog/source page; this page is the ecosystem routing page. Keep both and route between them rather than merging the source catalog into the ecosystem summary.
Routing Summary
Preferred order: Playwright for repeatable tests, browser-harness or agent-browser for visual QA, Browserbase/autobrowse when the task is a reusable site workflow, then Chrome DevTools MCP for live inspection. Source: Skill Resolver, 2026-06-27
Failure Modes
- Using Browserbase for a test that should become Playwright.
- Treating browser automation as docs/search.
- Forgetting to save the learned workflow as a skill.
- Leaving browser state unaudited after an agent run.
Verification Path
For saved browser workflows, replay the skill against the target site and capture the result. For QA, include screenshot or DOM evidence plus console/network status. Clean browser artifacts after the run. Source: Browserbase Browse.sh (Open Web Catalog), 2026-06-27; Skill Resolver, 2026-06-27
Timeline
- 2026-06-27 | Canonical ecosystem page created to separate Browserbase from Playwright, localhost testing, and agent search. Source: User request, 2026-06-27; Ecosystem Capability Coverage