Happy slice of mcp pizzaMCP.pizza

agentcontrolplane

MCP.Pizza Chef: humanlayer

Rather than something you add to a chat app, this sits on the other side of the relationship: a system that runs AI agents on its own schedule and hands them tools, including any MCP server you point it at. It can pause mid-job to ask a person to approve a step, or pass work to a second agent. It never left alpha, and there have been no changes since July 2025. Running it means Kubernetes, Docker and a paid model key.

Unmaintained · no changes since July 2025; never left alpha
Coding
Other

Use This MCP client To

Run an agent on a schedule without watching it Require my approval before an agent takes a risky step Give an agent only the outside tools I choose Hand part of a job to a second, specialised agent Ask a person a question mid-task and wait for the answer

README

Agent Control Plane (ACP)

ACP (Agent Control Plane) is a cloud-native orchestrator for AI Agents built on Kubernetes. It supports long-lived outer-loop agents that can process asynchronous execution of both LLM inference and long-running tool calls. It's designed for simplicity and gives strong durability and reliability guarantees. It embraces concepts from 12-factor-agents for agents that make asynchronous tool calls like contacting humans or delegating work to other agents.

⚠️ Note - ACP is in alpha.

Discord | Documentation | Examples

GitHub Repo stars License: Apache-2

Table of Contents

Architecture

agentcontrolplane FAQ

Is this project still maintained?
No. The last change was in July 2025, over a year ago, and it never moved past alpha. It has not been archived and no successor is named, though the same team's other projects remain very active.
Is this something I add to Claude or ChatGPT?
No, and this is the common mix-up. It works the other way round: it is the thing that runs agents and consumes MCP servers, so it belongs on a server, not in your chat app.
What do I need to run it?
A Kubernetes cluster, Docker, the kubectl command line, and a paid model key such as OpenAI's. A local cluster works for trying it out.
Can I use this to make an agent check with me first?
Yes. Human approval steps and treating a person as a tool the agent can call were both core features.
How hard is setup?
Firmly developer territory. Everything is defined in Kubernetes files applied from the command line.
Should I start a new project on it today?
Probably not. Alpha software with a year of silence behind it is a poor foundation; look for something under active development.