Write your MCP server + lifecycle hooks once, deploy to 29 AI-agent platforms — Claude Code, Codex, Cursor, Gemini, Copilot, OpenCode, Warp, Zed & more — from one defineConnector() API. Renders native config for every host (MCP · hooks · commands · skills) with built-in per-tool token telemetry.

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agent-connector": {
      "command": "npx",
      "args": [
        "-y",
        "ken-jo/agent-connector"
      ]
    }
  }
}

README Excerpt

> **If you BUILD an MCP integration:** write your server + hooks once with > `defineConnector()`, deploy it to every detected AI-agent platform, and measure > your own server's per-tool tokens. > **If you just USE agent CLIs:** run `agent-connector usage` to read their logs > and see per-CLI / per-model token totals — no connector, config, or install.

Tools (2)

detectstatus

Topics

ai-agentai-agentsclaude-codeclicodexcursordeveloper-toolsgemini-cligithub-copilothooksllmllm-toolsmcpmcp-servermiddleware

agent-connector

Quick Install

README Excerpt

Tools (2)

Topics

Related AI/LLM Servers