Hey Hunters 👋
Six months ago, I faced a problem that many of you might relate to:
Building AI agents is powerful, but every agent is an isolated island.
You build a coding agent → it can't talk to your data agent
You build a research agent → it can't collaborate with your automation agent
Want to run them in different clients? Rewrite everything.
Meanwhile, MCP (Model Context Protocol) was growing fast—1000+ tools in the ecosystem. But they were just tools, not agents. And there was no easy way to compose them into something more powerful.
That's when I realized: what if agents themselves could be MCP tools? So I built MCPC—a framework where:
✅ Agents are MCP tools (portable everywhere)
✅ Agents can discover and collaborate with each other
✅ You compose agents from existing MCP tools with just text descriptions
✅ True interoperability through standard MCP interfaces
Example: Build a "Codex fork" in one file:
- Describe what it does in plain English
- Reference MCP tools: desktop-commander, lsmcp, github
- Run it in Claude Desktop, VS Code, Cursor—anywhere
The best part?
Your agents become reusable building blocks. Someone else can compose your coding agent with their data agent to build something neither of us imagined.
This is agent interoperability through MCP—agents that discover, collaborate, and compose.(If you've heard of A2A: similar goal of agent interoperability, but MCPC uses the existing MCP standard. This means you can leverage 1000+ existing MCP tools immediately AND works in all MCP clients today—Claude Desktop, VS Code, Cursor, etc. No waiting for new client support!)

If you've ever wanted to:

True agent interoperability: Your agents discover and collaborate with each other automatically
Write once, run everywhere: Same agent code works in Claude, VS Code, Cursor—no rewrites
Compose from thousands of MCP tools: Build powerful agents by composing the growing MCP ecosystem

I'd love your feedback! 🚀

Hey Hunters 👋
Six months ago, I faced a problem that many of you might relate to:
Building AI agents is powerful, but every agent is an isolated island.
You build a coding agent → it can't talk to your data agent
You build a research agent → it can't collaborate with your automation agent
Want to run them in different clients? Rewrite everything.
Meanwhile, MCP (Model Context Protocol) was growing fast—1000+ tools in the ecosystem. But they were just tools, not agents. And there was no easy way to compose them into something more powerful.
That's when I realized: what if agents themselves could be MCP tools? So I built MCPC—a framework where:
✅ Agents are MCP tools (portable everywhere)
✅ Agents can discover and collaborate with each other
✅ You compose agents from existing MCP tools with just text descriptions
✅ True interoperability through standard MCP interfaces
Example: Build a "Codex fork" in one file:
- Describe what it does in plain English
- Reference MCP tools: desktop-commander, lsmcp, github
- Run it in Claude Desktop, VS Code, Cursor—anywhere
The best part?
Your agents become reusable building blocks. Someone else can compose your coding agent with their data agent to build something neither of us imagined.
This is agent interoperability through MCP—agents that discover, collaborate, and compose.(If you've heard of A2A: similar goal of agent interoperability, but MCPC uses the existing MCP standard. This means you can leverage 1000+ existing MCP tools immediately AND works in all MCP clients today—Claude Desktop, VS Code, Cursor, etc. No waiting for new client support!)

If you've ever wanted to:

True agent interoperability: Your agents discover and collaborate with each other automatically
Write once, run everywhere: Same agent code works in Claude, VS Code, Cursor—no rewrites
Compose from thousands of MCP tools: Build powerful agents by composing the growing MCP ecosystem

I'd love your feedback! 🚀

MCPC

Build your own Codex with 1000s MCP tools, use anywhere

Build your own Codex with 1000s MCP tools, use anywhere