Central authority for Claude Code Model Context Protocol (MCP) integration. Covers MCP server installation (HTTP, SSE, stdio transports), server management (add, list, remove), installation scopes (local, project, user), plugin-provided MCP servers, enterprise MCP configuration, MCP resources and @ mentions, MCP prompts as slash commands, OAuth authentication, environment variable expansion, Claude Code as MCP server, output limits, and MCP security. Assists with connecting external tools, configuring MCP servers, managing authentication, and troubleshooting MCP issues. Delegates 100% to docs-management skill for official documentation.
Manages MCP server installation, configuration, and authentication. Use when adding servers (HTTP/SSE/stdio), managing scopes (local/project/user), or troubleshooting connections.
/plugin marketplace add melodic-software/claude-code-plugins/plugin install claude-code-observability@melodic-softwareThis skill is limited to using the following tools:
references/audit-framework.mdreferences/credential-patterns.mdSTOP - Before providing ANY response about MCP (Model Context Protocol):
- INVOKE
docs-managementskill- QUERY for the user's specific topic
- BASE all responses EXCLUSIVELY on official documentation loaded
Skipping this step results in outdated or incorrect information.
Before responding, verify:
If ANY checkbox is unchecked, STOP and invoke docs-management first.
Central authority for Claude Code Model Context Protocol (MCP) integration. This skill uses 100% delegation to docs-management - it contains NO duplicated official documentation.
Architecture: Pure delegation with keyword registry. All official documentation is accessed via docs-management skill queries.
Keywords: MCP, Model Context Protocol, MCP servers, MCP tools, MCP resources, MCP prompts, HTTP transport, SSE transport, stdio transport, MCP installation, MCP scopes, local scope, project scope, user scope, .mcp.json, plugin MCP servers, enterprise MCP, managed-mcp.json, OAuth authentication, MCP resources, @ mentions, mcp__server__prompt, MCP output limits, MAX_MCP_OUTPUT_TOKENS, claude mcp serve
Use this skill when:
Use these keywords when querying docs-management skill for official documentation:
| Topic | Keywords |
|---|---|
| Overview | "MCP", "Model Context Protocol", "MCP overview" |
| What MCP Does | "what you can do with MCP", "MCP capabilities" |
| Popular Servers | "popular MCP servers", "MCP servers table" |
| Topic | Keywords |
|---|---|
| Installation Overview | "installing MCP servers", "MCP server configuration" |
| HTTP Transport | "HTTP transport MCP", "remote HTTP server", "--transport http" |
| SSE Transport | "SSE transport MCP", "Server-Sent Events transport" |
| Stdio Transport | "stdio transport MCP", "local stdio server" |
| Windows Setup | "Windows MCP servers", "cmd /c npx" |
| Topic | Keywords |
|---|---|
| Commands | "claude mcp add", "claude mcp list", "claude mcp remove", "/mcp command" |
| JSON Configuration | "add-json MCP", "MCP JSON configuration" |
| Import from Desktop | "add-from-claude-desktop", "import MCP servers" |
| Topic | Keywords |
|---|---|
| Scope Overview | "MCP installation scopes", "MCP scope levels" |
| Local Scope | "local scope MCP", "project-specific user settings" |
| Project Scope | "project scope MCP", ".mcp.json file" |
| User Scope | "user scope MCP", "cross-project MCP" |
| Scope Precedence | "MCP scope precedence", "scope hierarchy" |
| Topic | Keywords |
|---|---|
| Plugin MCP Servers | "plugin-provided MCP servers", "plugin MCP configuration" |
| Plugin Features | "CLAUDE_PLUGIN_ROOT", "plugin MCP lifecycle" |
| Topic | Keywords |
|---|---|
| Enterprise Setup | "enterprise MCP configuration", "managed-mcp.json" |
| Allowlists/Denylists | "allowedMcpServers", "deniedMcpServers", "MCP restrictions" |
| Enterprise Paths | "enterprise MCP paths", "managed MCP locations" |
| Topic | Keywords |
|---|---|
| MCP Resources | "MCP resources", "@ mentions MCP", "reference MCP resources" |
| MCP Prompts | "MCP prompts", "mcp__servername__promptname", "MCP slash commands" |
| Topic | Keywords |
|---|---|
| OAuth Setup | "MCP OAuth", "authenticate remote MCP", "/mcp authentication" |
| Token Management | "MCP token refresh", "clear MCP authentication" |
| Topic | Keywords |
|---|---|
| Environment Variables | "environment variable expansion", ".mcp.json variables", "${VAR}" |
| Output Limits | "MCP output limits", "MAX_MCP_OUTPUT_TOKENS", "output warning" |
| Claude as Server | "claude mcp serve", "Claude Code as MCP server" |
| Topic | Keywords |
|---|---|
| MCP Security | "MCP security", "MCP prompt injection", "trust MCP servers" |
What do you want to do?
claude mcp add (with transport flags)claude mcp list (show all servers)claude mcp get (server details)claude mcp remove (delete server)/mcp command (in-session status)claude mcp add-json (JSON configuration)claude mcp add-from-claude-desktop (import)claude mcp reset-project-choices (approval reset)claude mcp serve commandUser asks: "How do I add an MCP server?"
1. Invoke docs-management skill
2. Use keywords: "installing MCP servers", "claude mcp add"
3. Load official documentation
4. Provide guidance based EXCLUSIVELY on official docs
User asks: "I want to share MCP servers with my team and authenticate with OAuth"
1. Invoke docs-management skill with multiple queries:
- "project scope MCP", ".mcp.json file"
- "MCP OAuth", "/mcp authentication"
2. Synthesize guidance from official documentation
User reports: "My MCP server won't connect on Windows"
1. Invoke docs-management skill
2. Use keywords: "Windows MCP servers", "cmd /c npx"
3. Check official docs for Windows-specific requirements
4. Guide user through Windows setup from official docs
| Issue | Keywords for docs-management |
|---|---|
| Server won't connect | "installing MCP servers", specific transport type |
| Windows connection errors | "Windows MCP servers", "cmd /c" |
| Authentication failing | "MCP OAuth", "/mcp authentication" |
| Scope not working | "MCP installation scopes", "scope precedence" |
| Environment variables not expanding | "environment variable expansion", ".mcp.json variables" |
| Output too large | "MCP output limits", "MAX_MCP_OUTPUT_TOKENS" |
| Plugin MCP not loading | "plugin-provided MCP servers", "plugin MCP lifecycle" |
| Enterprise restrictions | "allowedMcpServers", "deniedMcpServers" |
This repository uses MCP servers for:
MCP servers are configured in the project's settings and should follow the patterns documented in official Claude Code documentation.
This skill provides the validation criteria used by the mcp-auditor agent for formal audits.
ALWAYS run claude mcp list FIRST before searching for configuration files. This is mandatory because:
~/.claude.json (NOT ~/.claude/.mcp.json - this path does not exist)mcpServers key may be deep in the file (line 200+) - easy to miss with partial reads# Step 1: Get authoritative list of configured servers
claude mcp list
# Step 2: Get details on specific server (shows scope and config)
claude mcp get <server-name>
Common Discovery Mistakes:
| Mistake | Why It Fails | Correct Approach |
|---|---|---|
Only searching for .mcp.json files | User MCP is in ~/.claude.json, not a separate file | Use CLI first, then verify files |
Reading only first N lines of ~/.claude.json | mcpServers may be deep in the file (line 200+) | Grep for mcpServers key first |
| Ignoring CLI discovery | Files may not reflect runtime state | Always start with claude mcp list |
Wrong path ~/.claude/.mcp.json | This path does not exist; user config is ~/.claude.json | Check official docs for current paths |
| Resource | Location | Purpose |
|---|---|---|
| Audit Framework | references/audit-framework.md | Query guides and scoring criteria |
| Category | Points | Key Criteria |
|---|---|---|
| Configuration Structure | 25 | Valid JSON, required fields |
| Server Entries | 25 | Valid server configs, proper format |
| Transport Config | 20 | Valid transport types, correct settings |
| Authentication | 15 | Proper auth, no exposed secrets |
| Scope Compliance | 15 | Correct scope (project/user/plugin) |
Thresholds: 85+ = PASS, 70-84 = PASS WITH WARNINGS, <70 = FAIL
The mcp-auditor agent (Haiku model) performs formal audits using this skill:
skills: mcp-integration/audit-mcp commandWhen auditing MCP configurations that reference external technologies (scripts, packages, runtimes), the auditor MUST validate claims using MCP servers before flagging findings.
Technologies Requiring MCP Validation:
Validation Rule:
Never flag a technology usage as incorrect without first:
Stale Data Warning:
Reference files:
Official Documentation (via docs-management skill):
Repository-Specific:
.claude/settings.jsonv1.1.0 (2025-12-25): Audit framework enhancement
v1.0.0 (2025-11-26): Initial release
Date: 2025-12-25 Model: claude-opus-4-5-20251101
This skill should be used when the user asks to "create a slash command", "add a command", "write a custom command", "define command arguments", "use command frontmatter", "organize commands", "create command with file references", "interactive command", "use AskUserQuestion in command", or needs guidance on slash command structure, YAML frontmatter fields, dynamic arguments, bash execution in commands, user interaction patterns, or command development best practices for Claude Code.
This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.
This skill should be used when the user asks to "create a hook", "add a PreToolUse/PostToolUse/Stop hook", "validate tool use", "implement prompt-based hooks", "use ${CLAUDE_PLUGIN_ROOT}", "set up event-driven automation", "block dangerous commands", or mentions hook events (PreToolUse, PostToolUse, Stop, SubagentStop, SessionStart, SessionEnd, UserPromptSubmit, PreCompact, Notification). Provides comprehensive guidance for creating and implementing Claude Code plugin hooks with focus on advanced prompt-based hooks API.