mcp-config

Generate MCP Configuration

Set up MCP server connections so Claude Code has live access to your Elasticsearch cluster and Kibana instance while developing plugins. Ask the user for:

1. Which MCP servers? Options:

   • Elasticsearch MCP only (recommended minimum)
   • Elasticsearch + Kibana MCP (recommended for full-stack plugin dev)
   • Elasticsearch + Kibana Agent Builder MCP (for Kibana 9.2+ / Serverless)

2. Elasticsearch connection:

   • URL (default: https://localhost:9200)
   • Auth method: API key, username/password, or cloud ID
   • SSL verification (default: true)

3. Kibana connection (if selected):

   • URL (default: https://localhost:5601)
   • Auth method: API key or username/password
   • For Agent Builder: Kibana Cloud URL

4. Docker or local?

   • Docker (recommended — no local install needed)
   • Local Node.js (for Kibana MCP community server)

5. Scope:

   • Project-level (.mcp.json in project root)
   • Global (~/.claude/mcp.json)

Generation Steps

1. Collect all connection details
2. Generate .mcp.json with the selected MCP servers
3. Generate a .env file for secrets (if not using env vars already)
4. Add .env to .gitignore if not already present
5. Test the connection by describing how to verify with claude /mcp

Important Rules

• Never hardcode API keys or passwords in .mcp.json — always use environment variable references
• Always recommend read-only API keys for development
• Remind the user to add .env to .gitignore
• For Docker-based ES MCP, use --network host when ES runs on localhost
• For the Kibana community MCP server, verify the user has cloned and installed it
• Test connectivity: after generating, tell the user to run claude /mcp to verify servers are connected
• Recommend starting with ES MCP alone — it covers 80% of use cases

Template: ES MCP Only (Docker)

{
  "mcpServers": {
    "elasticsearch": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "--network", "host",
        "-e", "ES_URL",
        "-e", "ES_API_KEY",
        "docker.elastic.co/mcp/elasticsearch"
      ],
      "env": {
        "ES_URL": "${ELASTICSEARCH_URL}",
        "ES_API_KEY": "${ELASTICSEARCH_API_KEY}"
      }
    }
  }
}