mcp-manager

rysweet's avatarfrom rysweet

Conversational interface for managing MCP (Model Context Protocol) server configurations in Claude Code

⭐10stars🔀3forks📁View on GitHub🕐Updated Dec 21, 2025⬇️1downloads

When & Why to Use This Skill

The MCP Manager skill provides a powerful conversational interface for managing Model Context Protocol (MCP) server configurations within Claude Code. It simplifies the developer workflow by allowing users to add, enable, disable, and validate MCP servers using natural language commands, effectively removing the need for manual JSON configuration editing and streamlining the integration of external tools and data sources.

Use Cases

  • Scenario 1: Quickly toggling specialized tools (like Puppeteer for browser automation or Filesystem for local file access) on and off during a development session to manage agent capabilities.
  • Scenario 2: Interactively configuring new MCP servers for databases or internal APIs by providing connection strings and environment variables through a guided chat interface.
  • Scenario 3: Troubleshooting integration issues by using the 'validate' command to check all MCP configurations for errors or connectivity problems in real-time.
  • Scenario 4: Managing complex development environments by exporting and importing MCP server configurations to ensure consistency across different machines or team members.
namemcp-manager
descriptionConversational interface for managing MCP (Model Context Protocol) server configurations in Claude Code
typeskill
activationStrategylazy-aggressive
activationContextWindow3
persistenceThreshold20

MCP Manager Skill

Overview

Natural language interface to MCP Manager CLI tool for managing Model Context Protocol server configurations. Users interact conversationally: "enable the filesystem MCP", "add a database server", "show me all my MCPs".

Activation

Activates on MCP keywords within 3-message window or explicit invocation: /mcp-manager

Commands

1. List MCPs

Display all configured MCP servers with status.

  • "List all my MCPs" / "Show me my MCP servers"
  • CLI: python3 -m mcp-manager.cli list

2. Enable MCP

Activate a disabled MCP server.

  • "Enable the filesystem MCP" / "Turn on puppeteer"
  • CLI: python3 -m mcp-manager.cli enable <server-name>

3. Disable MCP

Deactivate an MCP server without removing it. Requires confirmation.

  • "Disable the puppeteer MCP" / "Turn off github"
  • CLI: python3 -m mcp-manager.cli disable <server-name>

4. Add MCP

Add new MCP server interactively (collects name, command, args, env vars).

  • "Add a new MCP server" / "Configure a database MCP"
  • CLI: python3 -m mcp-manager.cli add <name> <command> [args...] --env KEY=VALUE

5. Remove MCP

Delete MCP server configuration completely. Requires confirmation with warning.

  • "Remove the puppeteer MCP" / "Delete the old-server"
  • CLI: python3 -m mcp-manager.cli remove <server-name>

6. Show MCP

Display detailed information for specific MCP server.

  • "Show me the filesystem MCP" / "Details for github server"
  • CLI: python3 -m mcp-manager.cli show <server-name>

7. Validate MCPs

Check all MCP configurations for errors.

  • "Validate my MCP configuration" / "Check for MCP errors"
  • CLI: python3 -m mcp-manager.cli validate

8. Export MCPs

Export configurations to JSON file for backup.

  • "Export my MCP configuration" / "Back up my MCPs"
  • CLI: python3 -m mcp-manager.cli export [output-file]

9. Import MCPs

Import configurations from JSON file.

  • "Import MCPs from backup.json" / "Restore my MCPs"
  • CLI: python3 -m mcp-manager.cli import <input-file> [--merge]

Usage Examples

Example 1: List and Enable

User: "List all my MCPs"

[Executes: cd .claude/scenarios && python3 -m mcp-manager.cli list]

Response: "You have 3 MCP servers configured: ✓ filesystem (enabled) - Local filesystem access ✓ github (enabled) - GitHub API integration ✗ puppeteer (disabled) - Browser automation"

Follow-up: "Turn on puppeteer"

[Executes: python3 -m mcp-manager.cli enable puppeteer]

"✓ Successfully enabled 'puppeteer' MCP server."

Example 2: Add New Server

User: "Add a new MCP for database access"

Interactive Collection:

  1. Name: "postgres-local"
  2. Command: "node /opt/mcp-servers/postgres/index.js"
  3. Args: "--port 5432 --host localhost"
  4. Env: "DATABASE_URL=postgresql://localhost:5432/mydb"

[Executes: python3 -m mcp-manager.cli add postgres-local "node /opt/mcp-servers/postgres/index.js" --port 5432 --host localhost --env "DATABASE_URL=postgresql://localhost:5432/mydb"]

"✓ Successfully added 'postgres-local' MCP server. Server is currently disabled. Enable with: 'enable postgres-local'"

Example 3: Remove Server

User: "Remove the puppeteer MCP"

Confirmation: "⚠️ WARNING: You're about to remove 'puppeteer' MCP server. This will delete the configuration completely. This action cannot be undone. Are you sure? (yes/no)"

[User confirms: "yes"]

[Executes: python3 -m mcp-manager.cli remove puppeteer]

"✓ Successfully removed 'puppeteer' MCP server."

Tool Invocation

All commands execute from .claude/scenarios/ directory:

cd .claude/scenarios && python3 -m mcp-manager.cli <command> [args]

Key Commands:

  • list - List all MCPs
  • enable <name> - Enable server
  • disable <name> - Disable server
  • add <name> <cmd> [args...] --env KEY=VAL - Add server
  • remove <name> - Remove server
  • show <name> - Show details
  • validate - Validate all configurations
  • export [file] - Export to JSON
  • import <file> [--merge] - Import from JSON

Output Handling:

  • Success: Exit code 0, stdout with ✓ prefix
  • Error: Non-zero exit code, stderr with ❌ prefix
  • Redact sensitive info (tokens, passwords) in responses

Error Handling

Common Errors:

  1. CLI Not Found: Offer installation instructions (see README)
  2. Server Not Found: List available servers, suggest alternatives
  3. Server Already Exists: Suggest show/remove/rename
  4. Permission Denied: Check file/directory permissions
  5. Malformed settings.json: Validate JSON, offer backup restore
  6. Invalid Command: Show common command patterns

For detailed error scenarios and troubleshooting, see .claude/scenarios/mcp-manager/README.md

Best Practices

  • Always confirm destructive operations (disable, remove)
  • Validate server names before executing
  • Redact sensitive information in responses
  • Provide clear error messages with actionable next steps
  • Show current state before and after changes

See Also

  • Full documentation: .claude/scenarios/mcp-manager/README.md
  • Tool creation guide: .claude/scenarios/mcp-manager/HOW_TO_CREATE_YOUR_OWN.md
  • MCP Protocol: https://modelcontextprotocol.io/
  • Claude Code Settings: .claude/settings.json

Version: 1.0.0 | Updated: 2025-11-24 | Maintainer: amplihack team