MCP Tools

Stacks integrate with the Nexus MCP server, providing tools for AI assistants to access stack prompts.

Available Tools

get-stack

Retrieve a compiled stack prompt by ID or slug.

Parameters:

Name	Type	Required	Description
stackId	string	Yes	Stack ID or slug

Example:

// Using stack slug
get-stack: { stackId: "tanstack-start" }

// Using stack ID
get-stack: { stackId: "stk_tanstack_start" }

Response:

{
  "id": "stk_tanstack_start",
  "name": "TanStack Start",
  "slug": "tanstack-start",
  "description": "Full-stack React framework with file-based routing",
  "category": "fullstack",
  "tokenBudget": "standard",
  "tokenCount": 4523,
  "compiledPrompt": "# TanStack Start Stack\n\n## Core Concepts...",
  "learningStatus": "complete"
}

list-stacks

List available stacks with optional filtering.

Parameters:

Name	Type	Required	Description
category	string	No	Filter by category
layer	number	No	Filter by layer (0-3)
search	string	No	Search query
limit	number	No	Max results (default: 20)

Example:

// List all fullstack stacks
list-stacks: { category: "fullstack" }

// Search for React-related stacks
list-stacks: { search: "react" }

Response:

{
  "stacks": [
    {
      "id": "stk_tanstack_start",
      "name": "TanStack Start",
      "slug": "tanstack-start",
      "description": "Full-stack React framework",
      "category": "fullstack",
      "layer": 1,
      "tokenCount": 4523,
      "isStarter": true,
      "isFeatured": true
    }
  ],
  "total": 1
}

Usage Patterns

Starting a Project

User: "Help me start a new project using the Cloudflare Workers stack"

AI: [Calls get-stack with "cloudflare-workers"]
AI: "Based on the Cloudflare Workers stack, here's how to set up your project..."

Combining Stacks

User: "I need help with a fullstack app using TanStack Start and Drizzle ORM"

AI: [Calls get-stack with "tanstack-start"]
AI: [Calls get-stack with "drizzle-orm"]
AI: "I've loaded both stack contexts. Here's how to integrate them..."

Discovering Stacks

User: "What stacks are available for frontend development?"

AI: [Calls list-stacks with { category: "frontend" }]
AI: "Here are the available frontend stacks:
     - Vite + React
     - Svelte 5 + SvelteKit
     - Vue 3 + Nuxt"

Integration

Claude Desktop

The Nexus MCP server automatically includes stack tools. Configure in claude_desktop_config.json:

{
  "mcpServers": {
    "nexus": {
      "command": "npx",
      "args": ["-y", "nexus-mcp"],
      "env": {
        "NEXUS_API_KEY": "your-api-key"
      }
    }
  }
}

Cursor / VS Code

Add to your MCP configuration:

{
  "servers": {
    "nexus": {
      "command": "npx",
      "args": ["-y", "nexus-mcp"]
    }
  }
}

Token Usage

Stack prompts count against your token budget. The tokenCount field shows the compiled prompt size:

Budget	Typical Size
Minimal	1,500 - 2,500
Standard	4,000 - 6,000
Comprehensive	8,000 - 12,000

Error Handling

Common errors:

Code	Meaning	Solution
STACK_NOT_FOUND	Stack ID/slug doesn't exist	Check the ID or use list-stacks
STACK_NOT_PUBLIC	Stack is private	Use your own stacks or public ones
COMPILATION_PENDING	Stack not yet compiled	Wait for background job to complete

Best Practices

1. Cache stack prompts - Reuse within a conversation
2. Use slugs - More readable than IDs
3. Check token count - Avoid exceeding context limits
4. Combine wisely - Don't overload with too many stacks