- What: Moda launched its public platform powered by a multi-agent AI system.
- Funding: $7.5 million Seed round led by Pear VC and Jeffrey Katzenberg’s WndrCo.
- Technology: Built using LangChain’s Deep Agents and a custom WebGPU-powered editable canvas.
- Architecture: Employs a three-agent system featuring a triage step, main agent loop, and dynamic context loading.
Moda has officially emerged from stealth, announcing a $7.5 million funding round and the public launch of an AI-driven design platform that replaces traditional prompt-to-image generation with sophisticated "Deep Agents." By utilizing a multi-agent architecture built on LangChain’s latest framework, Moda allows non-designers to create and iterate on production-grade visuals through a reasoning-based system that understands layout, typography, and brand identity.
The funding round saw significant participation from Pear VC and WndrCo, the venture firm co-founded by media mogul Jeffrey Katzenberg. Other investors include founders and executives from tech giants such as Dropbox, Stripe, Segment, Google, and Scale AI. The capital injection is slated to accelerate the development of Moda’s agentic workflow, which focuses on providing an editable, layered canvas rather than the static "flat" images produced by most contemporary AI generators.
A Multi-Agent Approach to Professional Design
At the core of Moda’s platform is a complex multi-agent system designed to handle the nuances of professional design. According to a technical deep dive published by LangChain, Moda transitioned its infrastructure to use "Deep Agents," a move that allows for more robust reasoning and cleaner execution of design tasks.
The system is structured around three primary agents that work in concert:
- The Triage Step: A lightweight agent that evaluates user intent and determines the necessary resources or sub-tasks required.
- The Main Agent Loop: This serves as the primary engine for executing design changes and reasoning through visual hierarchies.
- Dynamic Context Loading: A specialized process that injects brand-specific assets, color palettes, and typography rules into the agent’s working memory as needed.
While the majority of the system has been migrated to the Deep Agents framework, Moda disclosed that its "Design Agent" is currently running on a legacy LangGraph loop. The company is reportedly in the process of migrating this final component to the new architecture to unify its technical stack and improve performance.
Beyond the Prompt: The Editable Canvas
One of Moda’s primary differentiators in the crowded AI creative space is its departure from "prompt-only" generators. Most AI image tools generate a single, uneditable pixel layer. In contrast, Moda utilizes a WebGPU-based canvas that supports real-world design elements like editable layers, vector-like typography, and precise layout controls.
According to Moda, the platform uses a custom Domain Specific Language (DSL) specifically for layout reasoning. This allows the AI agents to "understand" the spatial relationship between elements, ensuring that when a user asks for a change—such as moving a headline or adjusting a brand color—the agent makes the change within the existing design structure rather than regenerating the entire image from scratch.
"Moda’s AI agent deeply understands a company’s brand and visual language," the company stated in its launch announcement. This deep understanding is facilitated by the agent's ability to reason through the custom DSL, ensuring that every output remains consistent with established brand guidelines.
Impact on the Design Industry
The launch of Moda signals a shift in the AI industry from "creative toys" to "production-grade tools." For developers and enterprises, this move demonstrates the power of multi-agent systems in solving high-stakes, high-precision tasks that were previously thought to require a human-in-the-loop for every minor adjustment.
For non-designers—such as marketing managers, product leads, and small business owners—the impact is immediate. The platform provides a bridge between a rough idea and a professional-grade asset without the steep learning curve of tools like Adobe Illustrator or Figma.
"Moda building a WebGPU canvas you can actually edit, plus an agent that understands layout, typography, and color, feels like a better path than prompt-only generators or template libraries," noted industry observers on Product Hunt.
Technical Transparency and Reliability
To ensure the reliability required for production-grade software, Moda relies heavily on LangSmith for tracing and debugging its multi-agent loops. This allows the engineering team to monitor the reasoning steps of each agent in real-time, identifying where an agent might misinterpret a design rule or fail to load the correct brand context.
By maintaining full observability through LangSmith, Moda can refine the "triage" step to ensure that the main agent loop receives the most relevant information for each user request. This technical rigor is essential for an AI system intended to be used in professional settings where brand consistency is non-negotiable.
What's Next for Moda
Following the public launch at moda.app, the company plans to finalize the migration of its entire agentic stack to the Deep Agents framework. With $7.5 million in new capital, the team is expected to expand its library of supported design formats and further enhance the agent's ability to reason through complex visual hierarchies.
As the AI landscape moves toward "agentic" workflows, Moda’s success will serve as a bellwether for whether multi-agent systems can truly replace traditional design software for the average professional.