[Feature]: trace correlation hooks for external observability/replay systems

[TheEleventhAvatar]

🚀 Feature Request

Playwright already has excellent tracing/replay tooling internally.

One thing that feels increasingly interesting for agentic/stateful workflows is easier integration with external observability systems (Langfuse, OpenTelemetry, custom trace pipelines, etc.).

Would there be value in exposing lightweight trace-correlation hooks/APIs for:

• associating Playwright sessions with external trace IDs
• attaching browser execution metadata to distributed traces
• replaying frontend execution alongside backend spans/tool calls
• debugging cross-system divergence in complex workflows

Not suggesting Playwright become an observability platform itself — more about enabling richer external integrations.

Example

No response

Motivation

This feels especially relevant for:

• AI agents operating through browser environments
• multi-service E2E workflows
• non-deterministic state bugs
• replay-based debugging systems

[dgozman]

@TheEleventhAvatar Could you please be more specific? What kind of an API would you like to see? Code snippets with a proposed API would be great.

[TheEleventhAvatar]

A possible direction could be lightweight trace propagation hooks rather than full observability integrations directly inside Playwright.

For example:

type TraceContext = {
  traceId: string;
  spanId?: string;
  metadata?: Record<string, string>;
};

await browser.newContext({
  traceContext: {
    traceId: "trace_123",
    metadata: {
      workflow: "agent-pr-review",
      runId: "run_456",
    },
  },
});

Then expose lifecycle hooks/events that external systems can subscribe to:

page.on("traceevent", (event) => {
  // navigation
  // action
  // assertion
  // network event
});

or:

context.exportTrace({
  traceId,
  format: "otel",
});

The main idea is making it easier for external systems to correlate:

• browser execution
• backend spans
• agent tool calls
• workflow replay timelines

without Playwright itself needing to become an observability platform.

Interesting use cases:

• replaying agent/browser failures alongside backend traces
• identifying frontend/backend divergence points
• correlating flaky E2E failures across services
• debugging long-running agent workflows operating through browser environments

I was mostly thinking about lightweight interoperability primitives rather than large built-in integrations.

[dgozman]

@TheEleventhAvatar I believe Playwright has all the required events already, e.g. console, framenavigated, request, response, etc. All of these should be enough to produce whatever trace you'd like. In fact, the built-in tracing is actually listening to these events and recording them in some specific json format.

Overall, I'm not sure we can act on such a broad feature request. If you are building some system with external tracing for Playwright, please file targeted issues for anything specific that you are missing. This way, it will be much easier to manage and actually implement the missing bits. Thank you!