MCP Client Integration

How to register ensemble-mcp with each supported AI coding tool. The easiest method is the auto-installer (ensemble-mcp install), but manual registration is also documented below.

Auto-Detection with ensemble-mcp install

The installer detects AI tools by checking for known config directories:

# Auto-detect and register all found tools
ensemble-mcp install

# Register only specific tools
ensemble-mcp install --tools cursor,copilot

# Project-local registration (writes to project dir instead of global config)
ensemble-mcp install --local

# Preview without making changes
ensemble-mcp install --dry-run

The installer creates backups of existing config files (.bak extension) before modification.

OpenCode

Global Registration

Config file: ~/.config/opencode/config.json

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "ensemble": {
      "type": "local",
      "command": ["uvx", "ensemble-mcp"]
    }
  }
}

Project-Local Registration

Config file: config.json in project root

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "ensemble": {
      "type": "local",
      "command": ["uvx", "ensemble-mcp"]
    }
  }
}

Agent & Skill Directories

Skills use directory format: each skill is placed as <skill-name>/SKILL.md.

Claude Code (Claude Desktop)

Global Registration

Config file: ~/.claude/claude_desktop_config.json

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Project-Local Registration

Config file: .claude.json in project root

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Skill Directories

GitHub Copilot (VS Code)

Global Registration

Config file: ~/.vscode/mcp.json

{
  "servers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Project-Local Registration

Config file: .vscode/mcp.json in project root

{
  "servers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Cursor

Global Registration

Config file: ~/.cursor/mcp.json

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Project-Local Registration

Config file: .cursor/mcp.json in project root

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Skill Directories

Windsurf

Global Registration

Config file: ~/.windsurf/mcp.json

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Project-Local Registration

Config file: .windsurf/mcp.json in project root

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Devin CLI

Global Registration

Config file: ~/.devin/mcp.json

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Project-Local Registration

Config file: .devin/mcp.json in project root

{
  "mcpServers": {
    "ensemble": {
      "command": "uvx",
      "args": ["ensemble-mcp"]
    }
  }
}

Skill Directories

Manual Registration (Generic stdio)

For any MCP-compatible tool not listed above, register the server with stdio transport:

{
  "ensemble": {
    "command": "uvx",
    "args": ["ensemble-mcp"]
  }
}

Or if installed via pip:

{
  "ensemble": {
    "command": "ensemble-mcp"
  }
}

The server communicates exclusively over stdio — it reads MCP requests from stdin and writes responses to stdout. All diagnostic output (startup banner, logs) goes to stderr.

Agent and Skill Files

The installer can also copy bundled agent and skill files:

Agents (7-agent pipeline)

Bundled agent files: team-ensemble.md, team-scope.md, team-craft.md, team-forge.md, team-trace.md, team-lens.md, team-signal.md.

These are automatically copied during ensemble-mcp install. To copy separately:

ensemble-mcp add-agents          # global scope
ensemble-mcp add-agents --local  # project-local scope

Skills (workflow skill)

Bundled skill file: ensemble-mcp-workflow.md — teaches AI agents when and how to invoke ensemble-mcp tools.

ensemble-mcp add-skills          # project-local scope (default)
ensemble-mcp add-skills --global # global scope

Next Steps

• CLI Reference — all install/uninstall flags
• Tool Reference — tools available after registration
• Troubleshooting — registration issues