AI in Intent Architect

Intent Architect has deeply integrated AI capabilities that operate inside the architecture you've designed, guided by your models, your standards, and your patterns and technologies. These capabilities are aligned with industry standards but also unique to the platform. The intent is simple: describe the design of your system, run the Software Factory, and out the other side comes working software - perfectly architected, consistent, readable, and maintainable. This is what we call the "golden path".

Intent's AI presents a single chat interface where you work with your designs, describe features, and guide development. Behind the scenes, when coding tasks need to be delegated, the system automatically hands off to a dedicated coding agent. You don't need to think about this distinction—it just works.

Deterministic generation + AI

To achieve this, Intent Architect deploys both deterministic code generation and generative AI. Deterministic code generation rolls out the architecture, infrastructure, and boilerplate that flow from your design. When coding tasks need to be implemented, the AI automatically handles them—whether that's through your conversation or delegated in the background. These systems work together inside the Software Factory, leveraging:

• Virtual Codebase which allows the AI to find and inspect files that have not yet been applied to the codebase, and to delegate implementation work as needed.
• Context engineering ensures every agent turn is informed by the model, your instruction files, and the relevant slice of the codebase - so output conforms to your application's architecture, standards, and structure.

Model-first development

In your single chat interface, you work with Intent's designers - the source of truth from which all generated code flows. Rather than asking AI to write code directly, you can describe a feature, drop in a PRD or screenshot, and have the AI shape the model itself. When implementation work is needed, it happens automatically and transparently. Every change is staged in-memory and only persisted with your explicit approval, so model changes are transparent at every step of the way. Your interactions are backed by:

• Powerful Tooling that reads, analyzes, and modifies the designs of your system. The AI can also interact with you (e.g. to ask clarifying questions where the requirements are ambiguous) and handle implementation tasks automatically.
• Designer-specific context - the AI works from live snapshots of your designers and diagrams, layered with per-solution guidance (AGENTS.md, INTENT.md, and any instruction files under .agents/) so your naming conventions, architectural rules, and project knowledge are applied to every change.
• Plan mode - for larger or ambiguous changes, iterative markdown planning with clarifying questions and your explicit approval before any changes are made.

What ships out of the box

• A single, powerful chat interface - work with your designers, describe features, and guide development without worrying about which agent is handling which task.
• Pluggable providers - OpenAI, Anthropic, Azure OpenAI, Gemini, OpenRouter, Ollama, or any OpenAI-compatible endpoint. Bring your own key.
• A full toolbox - file ops, designer/model edits, build & test, plan-mode tools, and conversation primitives—all transparently available to the AI.
• Customisation options - author your own .agent.md files, drop in SKILL.md skills, and write project-wide instruction files to shape every turn.
• Attachments - drag-and-drop, paste, or open PRDs, screenshots, code files, and model elements directly into the chat as conversation context.
• Tool-call transparency - every read, create, update, and delete the AI performs is shown as a color-coded interactive chip you can click to navigate straight to the affected element.

---

Documentation

Getting started

• AI Configuration - connect to your AI provider (OpenAI, Anthropic, Azure OpenAI, Gemini, OpenRouter, Ollama, or any OpenAI-compatible endpoint), expose Intent as an MCP server, and add external MCP servers per solution.
• Using the AI Chat - get started with the single chat interface for working with your designs and implementation.

Customising agents and context

• Agent Context Loading - where Intent looks for agent definitions, instruction files, and skills. The .agents/ folder under your solution and the dotfile conventions inside an application's output (AGENTS.md, CLAUDE.md, .cursor/rules/, .github/instructions/, etc.).
• Custom Agents - author your own .agent.md files: pick a context, choose tools, and write the system prompt that defines the agent's behaviour.
• Agent Tools - every tool an agent can be wired up with: file ops, designer/model edits, build/test, planning, and conversation tools.

Connecting to MCP Servers

• External MCP Servers - give this solution's coding agents extra tools by wiring in external MCP servers (filesystem, GitHub, your own internal tools, etc.). Configuration is stored per-solution in .agents/mcp.json and supports both stdio (launch a local command) and http (call a remote endpoint) transports, with ${VAR} substitution for secrets pulled from your environment. Each server has its own enable/disable toggle and live connection status, so you can park entries without deleting them.

Note

External MCP server tools are surfaced to coding-context agents only.

Intent MCP

• Intent Architect as an MCP server lets external agents (Claude Code, Copilot, Cursor, etc.) drive Intent Architect's designers directly - no friction when an outside harness needs to change managed code.

---

At a glance

You want to…	Go to
Plug in your OpenAI / Anthropic / Azure key	AI Configuration → AI Providers
Use Intent from Claude Code / Copilot / Cursor	AI Configuration → Intent MCP
Add an MCP server (filesystem, GitHub, etc.) to your AI	AI Configuration → MCP Servers
Get started with the AI chat interface	Using the AI Chat
Drop a project-wide instruction file	Agent Context Loading → Instruction files
Customize how the AI works for your solution	Agent Context Loading → Agent definitions
See what capabilities the AI has available	Agent Tools