Atsign has expanded its AI Architect product with new tools for designing and governing enterprise AI systems. The update adds Model Context Protocol integration and native AI agent modelling.
The changes aim to move AI-assisted software work beyond code generation and into earlier stages of system design, where architecture, security controls and governance are defined. The additions are available immediately to developers and enterprise teams using the Atsign platform.
Architecture focus
A central part of the update is a live workflow between large language models and a visual architecture canvas. Through Model Context Protocol, or MCP, integration, developers can use AI assistants to generate an initial system blueprint from natural language prompts, then refine that design within the same environment.
MCP is an open-source standard created by Anthropic that allows AI models to connect with external tools, databases and file systems. In Atsign's product, this means the model can work from a live architecture definition rather than separate prompts or exported specifications.
The approach reflects a wider shift in software development as companies try to use generative AI across more of the engineering process. While coding assistants have become common, many organisations still handle architecture review, compliance checks and security design in separate stages, slowing delivery and creating rework.
Alongside the co-architecting feature, Atsign has introduced native AI agent nodes inside the visual modelling environment. These are designed to let teams define agent roles, access limits and interaction patterns as part of the architecture itself.
This is aimed at companies building multi-agent systems, where several AI agents handle different tasks and exchange data or instructions. Visual modelling can give development, architecture and security teams a clearer view of how those agents operate and what permissions they hold before software is deployed.
Atsign's broader pitch rests on its direct-trust architecture. The system is built to connect people, entities and AI agents with identity, control and policy built in, rather than relying on central servers for communications.
This matters in a market where security concerns are rising alongside interest in agentic AI. Companies are under pressure to speed up software delivery with AI tools, but they also face growing scrutiny over how those systems are governed, what data they can access and whether they introduce new attack paths.
Recent attention on AI-related cyber threats has added to that pressure. In that context, tools that combine architecture design, implementation workflows and governance in one process are likely to attract businesses trying to avoid adding security reviews at the end of development.
"Most AI development tools stop at code generation. But enterprise AI development does not begin with code, it begins with architecture, governance, security boundaries, and system behavior," said Aparna Rayasam, Chief Executive Officer, Atsign. "Atsign AI Architect brings that governed architecture directly into the AI-assisted development workflow, helping transform today's AI coding assistants into secure-by-design architectural collaborators."
From prompt to blueprint
Atsign said teams can describe a system in natural language and receive a structured visual design in response. One example was a request to build a CRM platform with customer workflows, analytics pipelines and restricted-access AI agents.
From there, developers and architects can adjust the design visually, ask for revisions and review it against security or implementation constraints. This reduces the need for manual JSON exports because the AI model can pull architecture definitions directly from the live blueprint.
The release also addresses a practical issue inside large engineering teams: coordination between different functions. In many organisations, architects, developers and security specialists still use separate tools and formats, creating gaps between design decisions and the code eventually written.
Atsign argues that combining visual modelling, AI-assisted generation and implementation workflows in one environment could reduce those gaps. It says the product is intended to improve consistency across distributed teams without changing ownership structures or existing workflows.
The company cited external validation from a Broadband-Testing report that reviewed AI Architect and the wider Atsign platform in the context of secure AI application development. According to Atsign, the report examined how the software could be used to create application blueprints for LLM-assisted development and looked at a customer case in which an application previously built over several weeks was recreated in one afternoon.
The market opportunity is tied to a broader enterprise search for ways to move experimental AI projects into production systems. Many businesses can prototype quickly with generative models, but they struggle when they need to define controls, validate system behaviour and show compliance teams how autonomous agents are meant to operate.
By placing more of that work at the design stage, Atsign is trying to position its software as part of the governance layer around enterprise AI development rather than as another coding assistant. The product is designed to let teams validate behaviour before deployment and set permissions at the architectural layer.
Those features may appeal most to larger organisations that want more structure around AI software design, especially as agent-based systems become more common. Atsign says its new tools are intended to replace opaque prompt chaining with explicit architectural structure.