Aicode Toolkit

Run the init command in your project root to set up the toolkit configuration files. For new projects, pass a name and project type. For existing projects, run without flags.

# For an existing project:
npx @agiflowai/aicode-toolkit init

# For a new project:
npx @agiflowai/aicode-toolkit init --name my-app --project-type monolith

Add all three MCP servers to your agent's config file. For Claude Code this is .mcp.json in the project root; for Cursor it is .cursor/mcp.json. Each server runs independently and can be enabled or disabled based on your workflow.

{
  "mcpServers": {
    "scaffold-mcp": {
      "command": "npx",
      "args": ["-y", "@agiflowai/scaffold-mcp", "mcp-serve", "--admin-enable"]
    },
    "architect-mcp": {
      "command": "npx",
      "args": ["-y", "@agiflowai/architect-mcp", "mcp-serve", "--admin-enable"]
    },
    "style-system": {
      "command": "npx",
      "args": ["-y", "@agiflowai/style-system", "mcp-serve"]
    }
  }
}

Ask your AI agent to list available boilerplates. This should trigger the list-boilerplates tool from scaffold-mcp and return the project's configured templates.

Before editing a file, ask the agent to check the design pattern for that file using get-file-design-pattern. After making changes, use review-code-change to validate against your RULES.yaml constraints.

When generating UI code, ask the agent to list available themes or get component visuals using the style-system tools. This ensures generated code uses your project's actual CSS classes and shared components.