Calling agents from public app

Hi @amit_lohe
Is there a solution for using AI agents in public apps?

Unsupported source for RetoolAIAgentInvokeQuery: public_app

1 Like

Hey @Qamar_Gulraiz - this purposefully doesn't work at the moment. I'd love to learn more about the use case you're building here; is it a public-facing chatbot, or are you just trying to trigger an Agent from an app?

1 Like

Hi @kent ,
Yes, we are working on a customer service chatbot linking our website's MCP to it. The Agent chat block is placed in an app and then the app is embedded as a public app on our website.

This is to allow customers interactions without the need to login. Hope that helps in answering the question.

Thanks

Hi!

We also have interest in being able to embed a chat/agent app in our webpage. Is there any update as to when this would be possible?

1 Like

Hey @ricardoguerreiro - thanks for reaching out. I don't have any concrete updates yet, but it's on our list. :+1:

Hey guys! does the recent update to external apps and deprecation of public links bring the possibility to embed these as chat bot into webapps?

Hi @ricardoguerreiro,

It looks like the ticket created by Kent is still in the works. So I do not believe you will be able to invoke Agents from an app that is made Public.

Also just to note, I wouldn't say we deprecated public links, we changed the UI and now you can make any app public regardless of plan level.

Once an app is made public, the shareable link will be different, but it will appear in the same UI box as a normal app link.


Yeah, embedding conversational agents in webapps is totally doable now. Retool's custom component framework makes it pretty straightforward.

For production chatbot embedding that handles the backend infrastructure (conversation history, context management, user sessions), drop-in components work well. Weavy's copilot component integrates with OpenAI, Claude, or Gemini and gives you a ready-to-ship chat interface in about 10 minutes. Handles all the WebSocket stuff and knowledge context without custom backend work.

What kind of agent are you trying to embed - customer support, internal tool, something else?