remoteEndpoint fails with Playwright run-server because browserName is not forwarded

[davzucky]

Summary

We want to use @playwright/cli from a development container while the browser runs in a separate Playwright service container. This is the same setup we already use successfully for E2E tests with @playwright/test and chromium.connect(...).

With @playwright/cli, configuring browser.remoteEndpoint against a Playwright run-server endpoint fails before opening a page.

Use Case

Our devcontainer has a Playwright service:

playwright:
  image: mcr.microsoft.com/playwright:v1.60.0-noble
  command: npx -y playwright@1.60.0 run-server --port 3456 --host 0.0.0.0

Our application containers can reach it at:

ws://playwright:3456/

This works with @playwright/test using chromium.connect(...). We would like playwright-cli to work the same way so agents/developers can run browser automation commands without installing or launching a browser in the development image.

Reproduction

Start a Playwright server:

docker run --rm --name playwright-server \
  -p 3456:3456 \
  mcr.microsoft.com/playwright:v1.60.0-noble \
  /bin/sh -c "npx -y playwright@1.60.0 run-server --port 3456 --host 0.0.0.0"

Use .playwright/cli.config.json:

{
  "browser": {
    "browserName": "chromium",
    "isolated": true,
    "remoteEndpoint": "ws://localhost:3456/",
    "launchOptions": {
      "channel": "chrome"
    }
  }
}

Run:

playwright-cli open https://example.com --config=.playwright/cli.config.json

Actual Behavior

The CLI fails with:

Cannot read properties of undefined (reading 'launch')

With debug output, the inner error is server-side:

TypeError: Cannot read properties of undefined (reading 'launch')
    at PlaywrightServer._initLaunchBrowserMode

Expected Behavior

playwright-cli should use browser.browserName from the config when connecting to a Playwright run-server endpoint, and should open the page successfully:

### Page
- Page URL: https://example.com/
- Page Title: Example Domain

Local Validation

We patched the bundled playwright-core locally to pass browserName when connecting to the remote endpoint. With that patch and matching client/server Playwright versions, the CLI successfully opened https://example.com and produced a snapshot.

This appears to be caused by the playwright-core implementation used by @playwright/cli, not by the thin playwright-cli wrapper itself. I will link the corresponding core issue here once filed.

[davzucky]

Filed the corresponding playwright-core issue here: microsoft/playwright#40814