🌟 Editor's Note
It’s that time of the year again: our State of Software Architecture Survey 2026 is now open! It’s an annual survey on how teams document, design, and govern their architectures. We’ll share the results with the software architecture community in early 2027, and we’d love to hear from you! It will take less than 10 minutes.
This issue’s theme is automation: helping architects make informed decisions and get those decisions into their architectures quickly, without the manual work. For example, going from code to architecture diagrams (and the model) automatically or updating model objects using our MCP, are two patterns we’ve seen a lot of interest in from our customers.
So we recently added a feature to our MCP that lets teams view and edit diagrams using Claude, Cursor, Copilot, and ChatGPT! We also released the IcePanel plugin, a set of skills that analyse your codebase or import existing diagrams and generate them into IcePanel 🔌
On the customer side, our latest case study is an interview with Marex, a global financial services platform, on how they use IcePanel.
🎨 How to generate diagrams using the IcePanel plugin
The new IcePanel plugin generates diagrams and models from your codebase using Claude Code or GitHub Copilot. Point an LLM at your code or images and import the result into an IcePanel landscape: diagrams, model objects, connections and more. It’s useful for:
Migrating diagrams from other tools into IcePanel
Onboarding onto new repos using IcePanel diagrams
Automating diagram generation for teams
Verifying changes in the architecture
This plugin has four skills to get data into IcePanel. Instead of manually diagramming and translating diagrams from one tool to another, you point Claude Code at your repo(s). It will read the source code, build a model (objects and connections), and create diagrams in IcePanel through the API.
We walk through an end-to-end example of how to use it.
👀 How Marex uses IcePanel
In our latest interview case study, we spoke with Marex about how they use IcePanel. Marex is a global financial services platform with 4,000 people across 13+ major offices. About 500 of them work in technology, and together they manage more than 600 systems.
We spoke with three of Marex’s Enterprise Architects: Nick Pullen, Victor Popescu and Adam Fidler. They look after how the architecture landscape fits together and how the rest of the business understands it.
🤓 Reads + finds
From Microservices to Agents: the Next Evolution of Distributed Systems. Microservices split work by capability or domain. Agents pick their path at runtime, so now you’re orchestrating decisions, not services. Read
The Missing Architecture for Agentic Software Development. Writing code got faster, so the bottleneck moved to review, governance and delivery. Bain’s argument is that you need to redesign the engineering system, not just add tools. Read
How Netflix Uses AI Agents for Performance Engineering. An agent pipeline finds performance anti-patterns and checks its own fixes with tests, a canary and telemetry before a human sees the PR. Read
Agents are great coders. Can we guide them to be great software architects, too? This post covers three fixes: a human in the loop, architecture rules as executable constraints, and design principles packaged as agent skills. Read
Stay chill,
