Can AI agents automate Playwright?
Microsoft · desktop · developer-toolsrpa
Playwright is automatable today through the first-party Playwright MCP server. This record is Microsoft's Playwright library and its official agent surfaces. It is not BrowserStack, Cypress Cloud, UiPath, or Power Automate. Install @playwright/mcp with npx @playwright/mcp@latest. The server drives Chromium, Firefox, WebKit, or Edge through structured accessibility snapshots, not vision models. Tools cover navigation, click and type, screenshots, tabs, network mocking, and storage state. browser_run_code_unsafe runs arbitrary JavaScript in the Playwright server process and is RCE-equivalent; enable it only for trusted MCP clients. Default transport is local stdio; --port starts HTTP at /mcp. The official CLI is npx playwright for tests, codegen, and reports. Microsoft also documents playwright-cli (@playwright/cli) as a token-efficient agent CLI with snapshot commands. Official libraries exist for TypeScript, Python, Java, and .NET. The product LICENSE is Apache License 2.0. No public HTTP management API was established. Computer-use viability is unassessed; MCP is not a computer-use probe. MCP and CLI both score 8; MCP is the best path because official docs position it for persistent agentic browser loops over page structure.
Best path today: mcp · Overall automatability: 8/10
Modalities
| Modality | Status | Detail |
|---|---|---|
| API | unknown | coverage: unknown |
| SDK | yes | typescriptpythonjavadotnet official: yes |
| MCP | yes | verdict: official |
| Integrations | unknown | |
| CLI | yes | |
| Extensibility | yes | playwright-test webhooks: unknown |
| Data access | export: html-report · import: unknown | |
| RPA / UI automation | unknown | web-dom No repeatable UI probe of a Playwright product console was run. Playwright itself is the browser-automation library. Agents should use the official MCP server or CLI rather than driving an IDE UI. Computer-use viability is a separate unassessed path.
|
| Computer use | unknown | measured verdicts only — "unknown" means not yet probed by us |
Scores
| api | mcp | integrations | cli | rpa | computer use | overall |
|---|---|---|---|---|---|---|
| — | 8 | — | 8 | — | — | 8 |
Evidence
| Fact | Tier | Source | Date |
|---|---|---|---|
homepage | declared | source | 2026-09-03 |
license | declared | source | 2026-09-03 |
platforms | declared | source | 2026-09-03 |
modalities.sdk.exists | declared | source | 2026-09-03 |
modalities.sdk.languages[0] | declared | source | 2026-09-03 |
modalities.sdk.languages[1] | declared | source | 2026-09-03 |
modalities.sdk.languages[2] | declared | source | 2026-09-03 |
modalities.sdk.languages[3] | declared | source | 2026-09-03 |
modalities.mcp.first_party | declared | source | 2026-09-03 |
modalities.cli.exists | declared | source | 2026-09-03 |
modalities.extensibility.scripting[0] | declared | source | 2026-09-03 |
modalities.data_access.export[0] | declared | source | 2026-09-03 |
modalities.agent_docs.llms_txt | scraped | source | 2026-09-03 |
verdict.scores.mcp | declared | source | 2026-09-03 |
verdict.scores.cli | declared | source | 2026-09-03 |
freshness.watch[0].url | declared | source | 2026-09-03 |
freshness.watch[1].url | declared | source | 2026-09-03 |
freshness.watch[2].url | declared | source | 2026-09-03 |
freshness.watch[3].url | declared | source | 2026-09-03 |
Related tools
Other products in this database that share a category with Playwright.
Last verified 2026-09-03 · volatility high · JSON record