Opal - Google Labs Killer NEW App

Google Labs’ Opal is a no-code workflow builder aimed at turning natural-language requests into working LLM “mini apps,” with built-in steps for web research, text generation, and media creation. The core shift is that the boundary between what people configure and what they code is getting thinner: Opal chains prompts and model/tool calls into a reusable sequence, then lets users inspect and remix each step.

Opal arrives as a public preview from Google Labs, a team behind products such as NotebookLM. The transcript traces how similar tools emerged from the broader LLM ecosystem—moving from early frameworks like LangChain toward more agentic, multi-step systems—and argues that Opal is Google’s latest entry into the same category as n8n and Lindy. Unlike fully “hardcore agent” platforms, Opal focuses on prototyping: users can quickly assemble workflows, test them, and later translate the underlying prompts into code if they want.

The workflow process starts either by remixing examples from Opal’s gallery or by describing what to build. Opal then maps the request into a set of steps, showing intermediate outputs along the way. In the demo, a “blog post generator” workflow runs through web search, outline creation, full post writing, and banner image generation. A console view reveals which models and tools were used—for example, research calling Gemini 2.5 flash and subsequent generation using Gemini 2.0 flash—along with the retrieved web pages and the produced outline.

A key feature is step-level editing. The user can change which model powers a step (switching between Gemini 2.5 flash and Gemini 2.5 Pro, for instance) and swap image generation backends. The demo specifically replaces the banner image generator from a Gemini 2.0 flash image generator to Imagen 4, then adjusts the prompt wiring so the blog topic is embedded in the image prompt. Opal also supports additional user inputs: the user adds a “reader persona” input, threads it into the research and writing prompts, and reruns the workflow to produce a blog post tailored to an intended audience—here, an IT worker focused on automation.

Beyond remixing, Opal can generate workflows from scratch. The transcript describes creating a literature-review tool for uploaded archive papers, where Opal asks for an archive paper URL and a literature review topic, then builds a node graph to fetch and search related material. The user can insert further steps—such as using “deep research” with Gemini 2.5 flash to extract author information—illustrating how Opal can be extended into more complex pipelines.

The preview is US-only, with the transcript noting that a VPN may be needed to access it. Still, the takeaway is clear: Opal packages Google’s model and tool ecosystem into a visual, inspectable workflow system that non-coders can use immediately, while developers can later extract prompts and implement the same logic in code. The transcript frames this as an early version likely to expand over time, similar to how NotebookLM gained features after launch.