Browser not consistently launching in multi-agent Playwright automation setup

[deepakkamboj]

I’m using Voltagent to build a multi-agent system for Playwright-based test automation (repo: https://github.com/deepakkamboj/volt-agent-playwright).

Sometimes the Playwright browser launches successfully, but most of the time, it does not launch or connect properly when triggered by an agent. This causes the workflow to hang or fail inconsistently.

What I'm Trying to Build

• A planner agent to split tasks based on user input.
• Specialized agents like NavigationAgent, InteractionAgent, CodeGenAgent for different steps.
• A single browser instance shared across agents to avoid launching multiple browsers.
• Proper usage of prompts/templates for reliable multi-agent orchestration.

Problems Encountered

• Browser not reliably launching or connecting.
• Unclear how to manage prompts effectively across multiple agents.
• Need advice on maintaining a shared Playwright browser instance across agents.

Steps to Reproduce

• Use multiple agents via Voltagent.
• Try launching browser via Playwright in agent tool calls.
• Observe that browser sometimes starts but often fails silently.

Expected Behavior

• Browser should launch reliably when needed.
• Ability to share a browser session across different agents.

Would appreciate any guidance or example patterns to structure a reliable multi-agent system using Voltagent for real-world Playwright automation.

Thanks and regards,
Deepak Kamboj

[omeraplak]

Hey Deepak,
Thanks a lot for opening the issue and sharing all the details! We’re looking into it and will get back to you shortly.

Really excited to see what you’re building with VoltAgent!

[omeraplak]

Hey @deepakkamboj ,
Thanks a lot for the issue. I took a look at the problem and the use case. We’re planning to add an API that lets agents hold some context directly, which should help with cases like this. I’ll keep you posted soon!

[deepakkamboj]

Thank you so much for looking into the issue. I’d truly appreciate any guidance you can share on implementing end-to-end test automation using VoltAgent. You're also very welcome to contribute to the ongoing effort of building a multi-agent system for Playwright-based test automation. The repository can be found here: https://github.com/deepakkamboj/volt-agent-playwright.

[omeraplak]

Thanks! I’m working on an example for you. I’ll let you know as soon as it’s released.

In the meantime, you can follow the progress here:
https://github.com/VoltAgent/voltagent/pull/71/files

[deepakkamboj]

Thank you for working on this example — I truly appreciate your efforts. This has the potential to solve a significant challenge in test automation. If possible, please add me as a contributor to your repository so I can actively support and contribute to this initiative as well.

[omeraplak]

Hey @deepakkamboj ,
We’ve just released context support! Please check out the docs and the example below. If you have any more questions after that, could you create a new issue?

Docs: https://voltagent.dev/docs/agents/context/
Example: https://github.com/VoltAgent/voltagent/tree/main/examples/with-playwright

Also, don’t forget to run npm run volt update to get the latest changes!