Editorial illustration for Moda assesses migration of Design Agent from custom LangGraph loop
Moda's Design Agent Shifts to Modular AI Framework
Moda assesses migration of Design Agent from custom LangGraph loop
Moda’s engineering team has been busy turning experimental AI prototypes into tools that can actually ship. Their “Deep Agents” framework promises a more modular, scalable way to handle design‑focused tasks, yet the company still relies on an earlier codebase for its flagship Design Agent. That legacy component was built around a custom LangGraph loop, a pattern that predates the newer Deep Agents architecture.
While the newer agents follow a consistent structure—starting with a quick triage, moving into a main processing loop, pulling in context on the fly, and logging everything end‑to‑end—Moda’s developers are weighing the cost and benefit of bringing the Design Agent onto the same footing. The decision could affect everything from response latency to how easily the system can be debugged or extended. Understanding why the migration is on the table helps frame the broader push toward uniformity across Moda’s AI stack.
The Design Agent runs on a custom LangGraph loop -- an older implementation built before Deep Agents -- and the team is actively evaluating migrating it as well. All three agents share the same overall architecture: a lightweight triage step, a main agent loop, dynamic context loading, and full tracing in LangSmith. Context Engineering: The Details That Matter Getting a design agent to produce genuinely good output that is visually coherent and brand-accurate (not just technically correct) required a lot of intentional context engineering.
A Custom DSL Instead of Raw Scene Graph One of the hardest parts of building a design agent is figuring out how to represent visual layouts in a way LLMs can reason about effectively. Raw canvas state is verbose, coordinate-heavy, and token-expensive -- not a natural fit for how models think about structure and layout. Moda developed a context representation layer that gives the agent a cleaner, more compact view of what's on the canvas, which reduces token cost and improves output quality.
Is the migration worth the effort? The team’s assessment suggests that moving the Design Agent from its legacy LangGraph loop to the newer Deep Agents framework could streamline the shared architecture that already includes a lightweight triage step, a main loop, dynamic context loading and full tracing. Yet the older loop remains functional, and its performance characteristics have not been fully disclosed. Consequently, it is unclear whether the transition will yield measurable gains in latency or reliability, or merely add engineering overhead.
Because all three agents already operate under a unified pattern, a consolidated implementation might simplify observability through LangSmith and reduce maintenance complexity. Conversely, any shift could introduce unforeseen integration challenges, especially given the platform’s focus on delivering production‑grade designs to non‑designers.
For now, Moda continues to run its Design Agent on the custom loop while actively evaluating alternatives. The outcome of that evaluation will determine whether the platform’s AI‑driven design workflow gains any tangible advantage, or if the existing setup suffices for its target audience.
Further Reading
- LangChain and LangGraph Agent Frameworks Reach v1.0 Milestones - LangChain Blog
- Everything new in LangGraph v0.2 (2026). Full breakdown of features, breaking changes, code updates, and step-by-step migration tips. - Agent Framework Hub
- Best AI Agent Frameworks 2026: Developer Guide - AlphaCorp
- AI Agent Frameworks 2026: LangGraph vs CrewAI & More - Lets Data Science
Related Reading
Business & Startups OpenAI, a Series F San Francisco startup founded in 2015 by eight pioneers
Business & Startups Terminal-Bench 2.0 launches with Harbor, testing any container-installable agent
Business & Startups Zuckerberg Unveils Meta Compute to Build Global AI Infrastructure
From Archives Nvidia CEO clarifies DLSS 5 concerns in lengthy Lex Fridman interview
From Archives Apply: AI agents that reason, plan, execute workflows at Transform 2026
Common Questions Answered
What is the current implementation of Moda's Design Agent?
The Design Agent currently runs on a custom LangGraph loop, which is an older implementation predating the newer Deep Agents framework. This legacy component represents an earlier approach to building AI-driven design tools before Moda developed its more modular architecture.
Why is Moda considering migrating the Design Agent to the Deep Agents framework?
Moda is evaluating the migration to potentially streamline its architecture and align the Design Agent with the newer Deep Agents approach. The migration could potentially improve the agent's consistency, with benefits including a lightweight triage step, a main agent loop, dynamic context loading, and full tracing in LangSmith.
What shared architectural elements exist across Moda's AI agents?
All three of Moda's agents share a consistent architecture that includes a lightweight triage step, a main agent loop, dynamic context loading, and full tracing capabilities in LangSmith. This standardized approach allows for more modular and scalable AI-driven design tools.