Skip to content

Conversation

@lilac
Copy link
Contributor

@lilac lilac commented Dec 23, 2025

In v1 the MCP server for web automation only supports bridge mode. This change bring the deprecated MCP server back.

…core

- Move playwright from devDependencies to dependencies so it's available at runtime
- Add playwright-core to externals to prevent bundling issues
- Fixes 'Cannot find module playwright-core' error when running in http mode
…nshots

- Add waitForNetworkIdle method to BaseAgent interface in types.ts
- Call waitForNetworkIdle after action execution in tool-generator.ts
- Update PlaywrightAgent.waitForNetworkIdle to use optional timeout,
  respecting Playwright's configured navigationTimeout by default

This fixes an issue where clicking links that trigger navigation would
return screenshots of the old page before the new page finished loading.
Add common tools for browser control that execute directly without AI:
- navigate, reload, go_back, go_forward, get_url, get_title

Also add web_close tool to playwright-mcp and web-bridge-mcp.
Some MCP clients like Cursor perform client-side tool argument validation
that incorrectly rejected string values when the schema specified
union([string, number]). Simplified the value field to accept only strings,
which fixes compatibility issues while maintaining the same runtime behavior.

The transform() call is also removed since it's no longer needed.
@netlify
Copy link

netlify bot commented Dec 23, 2025

Deploy Preview for midscene ready!

Name Link
🔨 Latest commit 020b295
🔍 Latest deploy log https://app.netlify.com/projects/midscene/deploys/69611fe08a94cf00086dc94c
😎 Deploy Preview https://deploy-preview-1647--midscene.netlify.app
📱 Preview on mobile
Toggle QR Code...

QR Code

Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

Actions like KeyboardPress with no locate parameter should execute
directly without calling the AI model service. Previously, all actions
went through aiAction() which requires AI, causing 403 errors when
pressing keys like Enter/Tab without specifying an element to locate.

Changes:
- Add callActionInActionSpace to BaseAgent interface for direct execution
- Add actionNeedsAI() to detect if locate fields require AI resolution
- Use direct execution when no locate prompt is provided
- Fall back to aiAction only when AI reasoning is needed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant