MCP

The preset AI MCP server is how teams consume their design system in AI coding tools. The web dashboard is where you define your design system; MCP is where your team uses it.

Without MCP, assistants generate generic UI that drifts from your system. With MCP, assistants can:

• discover approved components, presets, tokens, and patterns
• validate proposals and generated code before apply
• follow compliance constraints by mode and scope

For the operator/designer/product framing — "what does MCP do for my team" — see presetai.dev/docs/platform/mcp.

Copy link to sectionThree layers

MCP connects three concerns:

• Canonical model — the source of truth: tokens, presets, patterns, recipes
• Policy and guardrails — what is allowed: forbidden primitives, context rules, maturity modes
• AI execution — what gets generated or changed: code, proposals, validations

Copy link to sectionRecommended usage pattern

For any AI session writing code against a preset-aware design system:

1. Read constraints and available artifacts
2. Map intent to approved presets and patterns
3. Validate the approach before generation
4. Generate output
5. Validate generated output
6. Propose or apply according to mode and approval policy

Copy link to sectionRead these next

• Tool categories — what kinds of tools exist and when to call them
• Read vs. write surfaces — safe usage boundaries
• Write lifecycle — propose → validate → apply
• Adaptive modes — how MCP behavior shifts with system maturity
• All MCP tools — auto-generated per-tool reference