Opal 2.0 by Google Labs - Now with smart agent, memory, routing and interactive chat
by•
Opal, no-code visual builder for AI workflows, just got a major upgrade — a new agent step that analyzes your goal, determines the best approach, and automatically calls the right tools, such as Veo for video or web search for research, to complete the task. Also, new tools to make it more capable — Memory, Dynamic Routing and Interactive Chat.
Replies
Huge upgrade to Google Opal today!
With Opal’s new agent step, workflows aren’t just linear automations anymore, they can actually analyze the goal, decide the best approach, and call the right tools (Veo for video, web search for research, etc.).
That’s a meaningful shift from “flow builder” to “decision-making system.”
The addition of Memory is huge. Persistent context is what turns demos into real products. Combine that with Dynamic Routing (@ Go to) and Interactive Chat, and you’re basically getting conditional logic + user-in-the-loop + long-term context, without code.
Example Opals you can build now:
🏠 An Interior Design Collaborator that remembers your home’s aesthetic.
📖 An Interactive Storybook creator that evolves with your choices.
💡 An Idea Generator that interviews you to refine your vision.
My take: this makes “Super Gems” on Gemini way more powerful because now your workflows can think before they act. If Google keeps tightening the loop between Gemini, Veo, and Opal, this could become one of the most underrated no-code AI builder ecosystems out there.
The app is weird.
I try to check app in the library and create my own and in both cases got error.
Looks the limits is poor.
BrandingStudio.ai
The dynamic routing piece is what catches my eye here, letting the agent decide which tool to call based on the goal rather than hardcoding the path is the right architecture for real-world workflows. Most no-code AI builders skip this and it shows.
Memory + Interactive Chat in the same loop is a meaningful combo. Does the memory layer persist across separate workflow runs, or is it session-scoped? That distinction matters a lot for use cases like client-facing automations.
Been building multi-model orchestration pipelines myself and the "smart routing" problem is genuinely hard to get right. Curious to see how Opal handles edge cases when the agent picks the wrong tool. Congrats on the launch!
This feels like the moment no-code tools stop being “flowcharts” and start acting like real assistants — memory + routing is what makes them actually usable day-to-day, not just demos.