AI-Assisted App Development One Playbook for Developers and AI Agents

Agent Skills are structured markdown files that load into your AI assistant's context (one per concern, per framework). They define correct import paths, component patterns, property bindings, and decision flows for Angular, React, Web Components, and Blazor. When a Skill is active, Copilot or Cursor stops relying on general training data, which is often outdated or wrong for your specific library version. Instead, it follows the Skill. The output is far more likely to compile, with correct imports and idiomatic patterns.

The CLI MCP (Model Context Protocol) server exposes Ignite UI's current documentation, API reference, and project-setup guides as queryable context for Angular, React, Web Components, and Blazor. Agents call it during a coding session to look up component APIs and follow CLI-backed setup and scaffolding steps; design tokens and theming come from the separate Theming MCP. Instead of reading a whole documentation site, the agent fetches the targeted, version-matched entries it needs. It updates with every CLI release.

Documentation is a text that an AI agent reads and interprets. The MCP server is a tool an agent calls and gets a structured response from. The difference in output quality is significant, especially at the edges of context limits, where text-heavy documentation degrades faster than targeted tool responses. Agent Skills add a second layer: explicit decision flows and patterns that prevent the agent from making the wrong choice before it even queries the API. Together, Skills and MCP replace approximation with a system.

Yes, and that's by design. A Skill is a markdown file. Edit it to match your team's conventions, add project-specific component patterns, enforce your architectural decisions. Version it alongside your codebase.

Both work independently and add value on their own. Skills alone significantly improve agent output for teams not yet running an MCP server. The MCP server alone gives agents live API access without Skills-level convention enforcement. Together they form the complete context stack.

Skills and the MCP server address hallucination at the source. Skills target the most common failure modes — wrong import paths, incorrect API signatures, outdated component names — by replacing training-data guesses with explicit instructions. The MCP server grounds the agent in the current library API, rather than a stale snapshot from its training data. The result is a significant reduction in the errors that cause rewrite cycles. No system eliminates hallucination entirely, but grounding the agent in both your conventions and the live library API gets you as close as current tooling allows.

Yes. The Theming MCP is in production and supports Angular, React, Web Components, and Blazor today.

The AI toolchain (Agent Skills, including generating an app from an image design, plus the CLI and Theming MCP servers) is built on top of Ignite UI, by Infragistics, which has shipped enterprise UI components for more than 35 years. The Skills are versioned alongside the library. The MCP server reflects the current release. There is no separate AI tier, no usage-based pricing, and no dependency on a third-party AI infrastructure that could change without notice.

Yes. Agent Skills work with GitHub Copilot (including VS Code Agent Mode), Cursor, Windsurf, Claude Code, Codex, Gemini CLI, and JetBrains Junie. Each has a persistent setup path — ig ai-config writes it for you once per project, and every session follows your standards from then on. The MCP servers work with any MCP client that supports STDIO.