Skip to content

Command Integration

MCP servers are automatically used by Claude Kit commands and modes. This guide shows which servers enhance each command and how they work together.

CommandMCP ServersEnhancement
/featureContext7, Sequential, Memory, Filesystem, PlaywrightFull development workflow with docs, planning, persistence
/fixSequential, Memory, Playwright, Context7, FilesystemSystematic debugging with browser testing
/testPlaywright, Filesystem, Context7, MemoryE2E tests, file management, testing patterns
/reviewPlaywright, Memory, Sequential, FilesystemVisual review, consistent standards
CommandMCP ServersEnhancement
/planSequential, Memory, Context7, FilesystemStructured breakdown, persistent decisions
/brainstormSequential, Memory, Context7Organized exploration, design persistence
/researchContext7, Sequential, MemoryReal-time docs, thorough analysis
CommandMCP ServersEnhancement
/indexFilesystem, MemoryProject scanning, structure persistence
ModePrimary MCPBest For
brainstormSequential + MemoryDesign sessions with persistent ideas
deep-researchSequential + Context7Thorough technical investigation
implementationFilesystem + Context7Focused coding with accurate docs
reviewPlaywright + MemoryUI review with consistent standards
orchestrationAll 5Complex multi-step parallel work
Terminal window
/feature Add user profile with avatar upload

MCP Flow:

  1. Context7 — Fetches React file upload documentation
  2. Sequential — Plans component structure step-by-step
  3. Memory — Recalls UI patterns from previous features
  4. Filesystem — Creates files in correct locations
  5. Playwright — Tests upload flow in browser
Terminal window
/fix Authentication fails intermittently on mobile

MCP Flow:

  1. Memory — Recalls auth architecture from previous sessions
  2. Sequential — Analyzes race conditions systematically
  3. Playwright — Tests on mobile device emulation
  4. Context7 — Checks session handling best practices
Terminal window
/research --mode=deep-research JWT vs Session authentication

MCP Flow:

  1. Sequential — Structures comparison step-by-step
  2. Context7 — Fetches current JWT and session library docs
  3. Memory — Stores findings for future reference
Terminal window
/index

MCP Flow:

  1. Filesystem — Scans project with directory_tree
  2. Memory — Stores structure as knowledge graph

Pattern: Research → Store → Recall

Session 1: /research React Server Components
→ Context7 fetches RSC documentation
→ Memory stores key patterns
Session 2: /feature Add data fetching
→ Memory recalls RSC patterns
→ Implementation uses correct approach

Pattern: Analyze → Decide → Remember

/brainstorm Database migration strategy
→ Sequential explores options step-by-step
→ Memory stores final decision
→ Future sessions recall the decision

Pattern: Test → Verify → Document

/test E2E checkout flow
→ Playwright automates browser testing
→ Context7 provides testing library docs
→ Tests use correct assertions

Pattern: Scan → Store → Navigate

/index
→ Filesystem scans project structure
→ Memory creates project knowledge graph
→ Future commands know where files are

Configuration differs by platform:

  • Linux/macOS: Use "command": "npx" directly
  • Windows: Use "command": "cmd" with "/c", "npx" in args

Examples below show Linux/macOS syntax. For Windows, see MCP Overview.

For basic usage, Context7 and Sequential are most impactful:

{
"mcpServers": {
"context7": {
"command": "npx",
"args": ["-y", "@upstash/context7-mcp"]
},
"sequential": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
}
}
}

For complete Claude Kit integration, use all 5 servers:

{
"mcpServers": {
"context7": {
"command": "npx",
"args": ["-y", "@upstash/context7-mcp"]
},
"sequential": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
},
"playwright": {
"command": "npx",
"args": ["-y", "@playwright/mcp"]
},
"memory": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-memory"]
},
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "."]
}
}
}

First run downloads npm packages (one-time). Subsequent starts are faster.

Servers start in parallel. All should be ready within seconds.

Each server runs as a separate process. Total overhead is minimal but consider disabling unused servers for resource-constrained environments.