SuperAPI AI Directory

next-devtools-mcp-skill

Agents & models

View repository

Updated 10-15 22:34

About this skill

Next.js DevTools MCP Skill

Skill by ara.so — Devtools Skills collection.

Expert skill for using next-devtools-mcp, a Model Context Protocol (MCP) server that provides Next.js development tools and utilities for AI coding agents. Provides runtime diagnostics, documentation search, browser automation, and development workflow automation for Next.js projects.

What is next-devtools-mcp?

next-devtools-mcp is an MCP server that gives AI coding agents powerful Next.js development capabilities:

  • Runtime Diagnostics (Next.js 16+): Query running Next.js applications for errors, routes, logs, and real-time state
  • Documentation Search: Access official Next.js docs and knowledge base directly in conversations
  • Browser Automation: Test and verify Next.js pages with Playwright integration
  • Development Workflows: Automated upgrade guides and feature enablement prompts

Installation

Quick Install (Recommended)

Install for all detected AI coding agents in your project:

npx add-mcp next-devtools-mcp@latest

Add -y to skip confirmation. Add -g for global installation across all projects.

Manual Installation

Add to your MCP client configuration file:

{
  "mcpServers": {
    "next-devtools": {
      "command": "npx",
      "args": ["-y", "next-devtools-mcp@latest"]
    }
  }
}

Agent-Specific Configuration

Claude Code / Claude Desktop:

claude mcp add next-devtools npx next-devtools-mcp@latest

Cursor: Click: Install in Cursor

Or manually: Cursor Settings → MCP → New MCP Server

Codex:

codex mcp add next-devtools -- npx next-devtools-mcp@latest

For Windows 11, update .codex/config.toml:

env = { SystemRoot="C:\\Windows", PROGRAMFILES="C:\\Program Files" }
startup_timeout_ms = 20_000

VS Code / Copilot:

code --add-mcp '{"name":"next-devtools","command":"npx","args":["-y","next-devtools-mcp@latest"]}'

Requirements

  • Node.js v20.19+ (latest maintenance LTS)
  • npm or pnpm
  • For runtime diagnostics: Next.js 16+ dev server running

Core Workflow

1. Always Initialize First

CRITICAL: Call the init tool at the start of every Next.js session:

Use the init tool to set up Next.js DevTools context

This establishes proper documentation requirements and context. For automation, add to your agent's configuration:

.claude/CLAUDE.md or .cursorrules:

When starting work on a Next.js project, ALWAYS call the `init` tool from
next-devtools-mcp FIRST to set up proper context and establish documentation
requirements. Do this automatically without being asked.

2. Start Next.js Dev Server (for Runtime Diagnostics)

For Next.js 16+ projects:

npm run dev
# or
pnpm dev

MCP is enabled by default at http://localhost:3000/_next/mcp. The next-devtools-mcp server auto-discovers and connects.

MCP Tools Reference

init

Initialize Next.js DevTools MCP context. Call this first in every session.

// Tool input schema
{
  project_path?: string  // Optional, defaults to current directory
}

Example prompts:

  • "Use the init tool to set up Next.js DevTools context"
  • "Initialize Next.js DevTools for this project"

What it does:

  • Sets up AI assistant context for Next.js development
  • Establishes documentation-first approach (use nextjs_docs for all queries)
  • Documents all available tools and workflows
  • Provides best practices checklist

nextjs_docs

Search and retrieve official Next.js documentation.

// Search for docs
{
  action: "search",
  query: string,           // e.g., "metadata", "generateStaticParams"
  routerType?: "app" | "pages" | "all"  // default: "all"
}

// Get full doc content
{
  action: "get",
  path: string,            // e.g., "/docs/app/api-reference/functions/refresh"
  anchor?: string          // e.g., "usage"
}

Example workflow:

Search Next.js docs for generateMetadata

Agent uses:

  1. nextjs_docs with action: "search", query: "generateMetadata"
  2. Reviews search results (titles, paths, snippets)
  3. nextjs_docs with action: "get", path: "/docs/app/api-reference/functions/generate-metadata"
  4. Provides answer based on official documentation

Common searches:

  • "Search Next.js docs for metadata generation"
  • "Find Next.js documentation on generateStaticParams"
  • "Look up middleware configuration in Next.js docs"
  • "Search App Router caching documentation"

nextjs_index

Query available runtime diagnostic resources from Next.js 16+ dev server.

// No input required
{}

Example prompts:

  • "What runtime diagnostics are available?"
  • "Show me what I can query from the dev server"
  • "List available Next.js MCP resources"

Output: List of available resource URIs (errors, routes, logs, etc.)

nextjs_call

Call a specific runtime diagnostic resource from Next.js 16+ dev server.

{
  uri: string  // Resource URI from nextjs_index
}

Example workflows:

Check for errors:

Next Devtools, what errors are in my Next.js application?

Agent calls: nextjs_call with uri: "nextjs://errors"

View route structure:

Next Devtools, show me the structure of my routes

Agent calls: nextjs_call with uri: "nextjs://routes"

Check dev server logs:

Next Devtools, what's in the development server logs?

Agent calls: nextjs_call with uri: "nextjs://logs"

Common URIs:

  • nextjs://errors - Runtime errors and warnings
  • nextjs://routes - App Router route structure
  • nextjs://logs - Development server logs
  • nextjs://config - Next.js configuration
  • nextjs://env - Environment variables

browser_eval

Automate browser testing with Playwright. Use for verifying pages, testing interactions, and detecting runtime issues.

// Start browser
{
  action: "start",
  browser?: "chrome" | "firefox" | "webkit" | "msedge",  // default: "chrome"
  headless?: boolean  // default: true
}

// Navigate to URL
{
  action: "navigate",
  url: string
}

// Click element
{
  action: "click",
  selector: string  // CSS selector
}

// Type text
{
  action: "type",
  selector: string,
  text: string
}

// Fill form
{
  action: "fill_form",
  fields: Array<{ selector: string; value: string }>
}

// Execute JavaScript
{
  action: "evaluate",
  script: string
}

// Take screenshot
{
  action: "screenshot",
  path?: string  // default: auto-generated
}

// Get console messages
{
  action: "console_messages"
}

// Close browser
{
  action: "close"
}

Example workflow - Test Next.js page:

Test the homepage at localhost:3000 and take a screenshot

Agent executes:

// 1. Start browser
{ action: "start", headless: true }

// 2. Navigate
{ action: "navigate", url: "http://localhost:3000" }

// 3. Wait for content
{ action: "evaluate", script: "document.querySelector('h1')?.textContent" }

// 4. Screenshot
{ action: "screenshot" }

// 5. Check for errors
{ action: "console_messages" }

// 6. Clean up
{ action: "close" }

Common patterns:

Verify upgrade (Next.js 15 → 16):

Verify that my app works on localhost:3000 after the upgrade

Test navigation:

Test navigating from home to /about page

Check for hydration errors:

Load the page and check for React hydration errors in the console

Important: For Next.js 16+, prefer nextjs_index and nextjs_call over browser_eval console messages for error detection.

Pre-Configured Prompts

upgrade-nextjs-16

Guide for upgrading from Next.js 15 to Next.js 16.

Help me upgrade my Next.js app to version 16

What it does:

  1. Analyzes current project setup
  2. Updates dependencies in package.json
  3. Applies codemods for breaking changes
  4. Updates configuration files
  5. Tests the upgrade with browser verification

enable-cache-components

Enable Cache Components mode in Next.js 16.

Enable Cache Components in my Next.js app

What it does:

  1. Adds cache: "components" to next.config.ts
  2. Migrates components to use "use cache" directive
  3. Updates data fetching patterns
  4. Provides examples and documentation

Knowledge Base Resources

The MCP server includes focused documentation resources automatically available to agents:

Cache Components (12 sections):

  • cache-components://overview
  • cache-components://core-mechanics
  • cache-components://public-caches
  • cache-components://private-caches
  • cache-components://runtime-prefetching
  • cache-components://request-apis
  • cache-components://cache-invalidation
  • cache-components://advanced-patterns
  • cache-components://build-behavior
  • cache-components://error-patterns
  • cache-components://test-patterns
  • cache-components://reference

Migration Guides:

  • nextjs16://migration/beta-to-stable
  • nextjs16://migration/examples

Fundamentals:

  • nextjs-fundamentals://use-client

Resources are loaded on-demand by the agent as needed.

Real-World Examples

Example 1: Debug Runtime Error

User request:

My Next.js app shows a blank page. Help me figure out what's wrong.

Agent workflow:

// 1. Initialize (if not done)
init({ project_path: "." })

// 2. Check runtime errors
nextjs_call({ uri: "nextjs://errors" })

// Output shows: "ReferenceError: window is not defined at Component.render"

// 3. Search docs for server component issues
nextjs_docs({ 
  action: "search", 
  query: "window is not defined server component" 
})

// 4. Get full doc content
nextjs_docs({ 
  action: "get", 
  path: "/docs/app/building-your-application/rendering/server-components"
})

// 5. Provide solution: Add "use client" directive

Example 2: Upgrade to Next.js 16 with Cache Components

User request:

Upgrade my app to Next.js 16 and enable Cache Components

Agent workflow:

// 1. Initialize
init({ project_path: "." })

// 2. Use upgrade prompt
// Triggers: upgrade-nextjs-16 prompt
// - Updates package.json
// - Runs codemods
// - Updates config

// 3. Enable Cache Components
// Triggers: enable-cache-components prompt
// - Adds cache: "components" to next.config.ts
// - Shows migration examples

// 4. Verify with browser
browser_eval({ action: "start" })
browser_eval({ action: "navigate", url: "http://localhost:3000" })
browser_eval({ action: "console_messages" })
browser_eval({ action: "screenshot" })
browser_eval({ action: "close" })

// 5. Check runtime state
nextjs_call({ uri: "nextjs://errors" })

Example 3: Implement Metadata Generation

User request:

Add proper SEO metadata to my blog post page

Agent workflow:

// 1. Search docs
nextjs_docs({ 
  action: "search", 
  query: "generateMetadata" 
})

// 2. Get detailed docs
nextjs_docs({ 
  action: "get", 
  path: "/docs/app/api-reference/functions/generate-metadata" 
})

// 3. Implement based on docs

Example implementation:

// app/blog/[slug]/page.tsx
import { Metadata } from 'next'

interface Props {
  params: { slug: string }
}

export async function generateMetadata({ params }: Props): Promise<Metadata> {
  const post = await fetch(`${process.env.API_URL}/posts/${params.slug}`)
    .then(res => res.json())
  
  return {
    title: post.title,
    description: post.excerpt,
    openGraph: {
      title: post.title,
      description: post.excerpt,
      images: [{ url: post.image }],
    },
  }
}

export default async function BlogPost({ params }: Props) {
  const post = await fetch(`${process.env.API_URL}/posts/${params.slug}`)
    .then(res => res.json())
  
  return (
    <article>
      <h1>{post.title}</h1>
      <div dangerouslySetInnerHTML={{ __html: post.content }} />
    </article>
  )
}

Example 4: Migrate to Cache Components

User request:

Convert my data fetching component to use Cache Components

Before (traditional approach):

// app/posts/page.tsx
export const revalidate = 3600

export default async function PostsPage() {
  const posts = await fetch('https://api.example.com/posts')
    .then(res => res.json())
  
  return (
    <div>
      {posts.map(post => (
        <article key={post.id}>
          <h2>{post.title}</h2>
        </article>
      ))}
    </div>
  )
}

Agent searches docs:

nextjs_docs({ 
  action: "search", 
  query: "use cache directive" 
})

// Then gets resource
// (agent has cache-components://core-mechanics available)

After (Cache Components):

// app/posts/page.tsx
import { Posts } from './Posts'

export default function PostsPage() {
  return <Posts />
}

// app/posts/Posts.tsx
"use cache"
export const revalidate = 3600

export async function Posts() {
  const posts = await fetch('https://api.example.com/posts')
    .then(res => res.json())
  
  return (
    <div>
      {posts.map(post => (
        <article key={post.id}>
          <h2>{post.title}</h2>
        </article>
      ))}
    </div>
  )
}

Configuration Patterns

next.config.ts with Cache Components

import type { NextConfig } from 'next'

const config: NextConfig = {
  experimental: {
    cache: 'components',  // Enable Cache Components
  },
}

export default config

Environment Variables

Use environment variables for API keys and sensitive data:

# .env.local
API_URL=https://api.example.com
NEXTAUTH_SECRET=your-secret-here
NEXTAUTH_URL=http://localhost:3000

Access in code:

const apiUrl = process.env.API_URL
const secret = process.env.NEXTAUTH_SECRET

MCP Auto-Discovery

Next.js 16+ automatically enables MCP at:

http://localhost:{PORT}/_next/mcp

No configuration needed. The next-devtools-mcp server discovers it automatically.

Troubleshooting

Issue: Runtime diagnostics not available

Symptoms: nextjs_index returns empty or nextjs_call fails

Solution:

  1. Verify Next.js 16+ is installed: npm list next
  2. Ensure dev server is running: npm run dev
  3. Check dev server logs for MCP endpoint
  4. Verify port (default 3000, check next.config.ts for custom port)

Issue: Documentation search returns no results

Symptoms: nextjs_docs with action: "search" returns empty

Solution:

  1. Try broader search terms (e.g., "metadata" instead of "generateMetadata function")
  2. Check spelling
  3. Try routerType: "all" to search both App and Pages Router docs
  4. Use multiple searches for complex topics

Issue: Browser automation fails on first run

Symptoms: browser_eval with action: "start" times out

Solution:

  1. Playwright installs browsers on first run (this is automatic but takes time)
  2. Increase timeout expectations for first use
  3. On Windows 11 with Codex, ensure environment variables are set in .codex/config.toml
  4. Run manually once: npx playwright install chromium

Issue: Init tool not being called

Symptoms: Agent doesn't use Next.js docs properly, context is missing

Solution:

  1. Explicitly prompt: "Use the init tool to set up Next.js DevTools context"
  2. Add auto-init instruction to agent config (see Core Workflow section)
  3. Verify MCP server is connected (check agent's MCP status)

Issue: Upgrade prompt makes incorrect changes

Symptoms: Breaking changes after running upgrade-nextjs-16 prompt

Solution:

  1. Always commit code before running upgrades
  2. Review changes made by codemods
  3. Use nextjs_docs to verify migration patterns
  4. Test with browser_eval before deploying
  5. Check nextjs://errors for runtime issues

Best Practices

1. Documentation-First Approach

Always use nextjs_docs before implementing Next.js features:

// Search first
nextjs_docs({ action: "search", query: "feature-name" })

// Get detailed docs
nextjs_docs({ action: "get", path: "/docs/..." })

// Then implement based on official documentation

2. Runtime Diagnostics Over Browser Console

For Next.js 16+ projects, prefer built-in diagnostics:

// ✅ Preferred
nextjs_call({ uri: "nextjs://errors" })

// ❌ Avoid when above is available
browser_eval({ action: "console_messages" })

Browser automation is still valuable for visual verification and interaction testing.

3. Initialize Every Session

Make init the first tool call:

// At start of every Next.js session
init({ project_path: "." })

Or automate with agent configuration (see Core Workflow).

4. Two-Step Documentation Workflow

Don't skip the search step:

// Step 1: Search to find relevant docs
nextjs_docs({ action: "search", query: "..." })

// Step 2: Get full content from search results
nextjs_docs({ action: "get", path: "..." })

This ensures you get the most relevant and up-to-date documentation.

5. Verify Changes with Multiple Tools

After making changes, use a combination:

// 1. Check runtime errors
nextjs_call({ uri: "nextjs://errors" })

// 2. Visual verification
browser_eval({ action: "start" })
browser_eval({ action: "navigate", url: "http://localhost:3000" })
browser_eval({ action: "screenshot" })
browser_eval({ action: "close" })

// 3. Check routes structure
nextjs_call({ uri: "nextjs://routes" })

Common User Prompts

Recognize and handle these user requests:

  • "Help me debug my Next.js app" → Use nextjs_call for errors
  • "Upgrade to Next.js 16" → Use upgrade-nextjs-16 prompt
  • "Enable Cache Components" → Use enable-cache-components prompt
  • "Search Next.js docs for X" → Use nextjs_docs two-step workflow
  • "Test my page at localhost:3000" → Use browser_eval workflow
  • "What errors are showing?" → Use nextjs_call with nextjs://errors
  • "Show my routes" → Use nextjs_call with nextjs://routes
  • "Check the dev server logs" → Use nextjs_call with nextjs://logs
  • "Add metadata to my page" → Search docs, then implement generateMetadata
  • "How do I configure X in Next.js?" → Use nextjs_docs to find config docs

Summary

The next-devtools-mcp skill enables AI coding agents to:

  1. Initialize properly with the init tool (critical first step)
  2. Access official Next.js documentation on-demand with nextjs_docs
  3. Query runtime diagnostics from Next.js 16+ dev servers with nextjs_index and nextjs_call
  4. Automate browser testing with browser_eval using Playwright
  5. Execute complex workflows like upgrades and migrations with pre-configured prompts
  6. Access knowledge base resources for Cache Components, migrations, and fundamentals

Always prioritize documentation lookup over assumptions, verify changes with runtime diagnostics, and test thoroughly with browser automation.