Model Context Protocol is becoming the universal API for AI coding assistants. Cursor, Claude Code, VS Code with Copilot, Windsurf — they all speak MCP. Axint ships a built-in MCP server that exposes five tools your AI agent can call.
The five tools
- axint_scaffold — generate a new Axint project with pinned deps, tsconfig, and a starter intent
- axint_compile — compile a TypeScript or Python intent to Swift + Info.plist + entitlements
- axint_validate — run the 4-stage validator without emitting files
- axint_list_templates — browse the built-in template library
- axint_template — fetch a specific template's source code
One-click install
We built native deeplinks for every major editor:
Cursor: Click the install button on axint.ai and Cursor opens with the MCP server pre-configured. No JSON editing.
VS Code: Same pattern — a vscode:// deeplink that registers the server.
Windsurf: Native deeplink support.
Claude Code / Claude Desktop: A single command:
claude mcp add axint -- npx -y @axintai/compiler mcpThat's it. Your AI agent can now scaffold, compile, and validate App Intents without you touching Xcode.
Why this matters
The future of software development is agents writing code for agents. An AI coding assistant in Cursor can read your Swift project, draft an intent in TypeScript, compile it with Axint, and open a PR — all without a human touching Xcode or a config file.
MCP is the protocol. App Intents is the target. Axint is the bridge.
Try it
Install the MCP server, open your editor, and ask your AI assistant: "Create an App Intent that searches my notes by keyword." Watch it scaffold, compile, and validate in seconds.