How Playwright connects to remote browsers.

Playwright normally launches browsers on the same machine. Remote operation changes that ownership boundary. A browser server owns the browser process. Your application owns the Playwright client.

Playwright's native connect method attaches remotely. It expects a Playwright WebSocket endpoint. The BrowserServer wsEndpoint method produces it. Another provider may supply an equivalent endpoint.

The endpoint must match Playwright's protocol. Any WebSocket address is insufficient. A Chrome debugging socket is different. A browser viewing URL is also different.

Connection success proves protocol compatibility. It does not prove task reliability. State, artifacts, timeouts, and cleanup still matter. Evaluate the whole operating contract.

Legba labels its current API public preview. Its endpoint reference remains under refinement. The organization-specific API host is still finalizing. Production decisions should reflect that status.

Current docs describe REST instance lifecycle. Teams can create isolated browser instances. They can list active instances. They can destroy completed instances.

Instance responses include an access_url. The docs describe browser access through that value. They never define a Playwright protocol. They never define a CDP protocol.

Therefore, current Playwright compatibility is undocumented. The access_url cannot establish compatibility alone. Wait for an explicit protocol contract. Recheck canonical docs before implementation.

Playwright exposes two relevant remote attachment methods. They do not offer equivalent behavior. Native connect speaks Playwright's protocol. connectOverCDP speaks Chrome DevTools Protocol.

Native connect requires compatible Playwright versions. Client and server major versions must match. Minor versions must also match. Patch versions may differ.

connectOverCDP supports Chromium browsers only. Official documentation describes significantly lower fidelity. Some Playwright capabilities may behave differently. Native protocol remains the stronger default.

A provider's REST URL is neither path automatically. The provider must name the transport. It must document authentication and versions. Otherwise, compatibility remains unknown.

The following pattern mirrors Playwright's documented method. It deliberately uses a generic environment variable. Replace it only with provider documentation. Never transform an unrelated access URL.

The client connects before creating a context. That context receives its own pages. The example closes context resources explicitly. It then closes the connected browser object.

This example is provider-neutral. It uses a placeholder endpoint. It was not tested against Legba. It does not prove Legba compatibility.

Production code needs provider-specific lifecycle handling. It also needs retries and artifact capture. Those details depend on published contracts. Keep them outside generic connection helpers.

Playwright browser contexts provide independent sessions. They isolate cookies and local storage. Multiple contexts can share one browser. Each context still needs explicit ownership.

Authentication state can reduce repeated login work. Playwright stores reusable state through storageState. That file may contain sensitive cookies. It may also contain sensitive headers.

Playwright recommends keeping state outside repositories. Add its directory to version-control exclusions. Delete expired state when appropriate. Never expose reusable state through client bundles.

Remote providers add another storage boundary. Ask where context data resides. Ask whether profiles persist after termination. Ask how deletion is verified. Document every answer before reuse.

Remote debugging needs durable evidence. A local stack trace may be insufficient. Browser actions happened elsewhere. Network timing and page state can disappear quickly.

Playwright tracing records browser operations. Trace Viewer exposes actions and metadata. Traces can include screenshots and snapshots. Their contents may include sensitive application data.

Playwright Test supports retry-focused tracing. Official guidance recommends tracing the first retry. That balances evidence against performance costs. Plain library usage requires explicit tracing calls.

Artifact transfer remains provider-specific. Ask where trace files are written. Ask whether disconnection preserves them. Ask when downloads become available. Test retrieval during partial failures.

Playwright exposes browser and context lifecycle methods. The browser close method clears connected contexts. It then disconnects the client. Close contexts first for orderly artifact flushing.

Provider infrastructure may have another lifecycle. A disconnected client may leave resources active. The provider's REST contract must own termination. Browser closure never proves provider termination.

Failures can occur between every cleanup step. Context closure can fail. Client disconnection can race network loss. Provider termination can also fail independently.

Use resource identifiers for reconciliation. List remaining resources after interruptions. Retry only documented safe operations. Confirm a final terminated state. Keep browser errors and lifecycle errors separate.

Begin with one production-shaped Playwright task. Include navigation and authentication. Include expected outputs and artifacts. Include a normal failure path.

Write compatibility requirements before choosing providers. Name the desired connection method. Name the supported browser engines. Name acceptable version constraints.

Legba currently supports lifecycle research only. Its preview docs lack Playwright protocol details. They also lack CDP endpoint details. Current evidence cannot justify implementation.

Return after published compatibility appears. Then test the exact documented endpoint. Pin compatible Playwright versions. Recheck documentation before each upgrade. Keep guessed endpoints out of production.