Agent Control Plane

v1.0.0Coding Agentsstable

ACP is the Agent Control Plane - a distributed agent scheduler optimized for simplicity, clarity, and control. It is designed for outer-loop agents that run without supervision, and make asynchronous tool calls like requesting human feedback on key o

agent-control-planeagentsframeworkhuman-in-the-loopkubernetes
Share:
409
Stars
0
Downloads
0
Weekly
0/5

What is Agent Control Plane?

Agent Control Plane is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to acp is the agent control plane - a distributed agent scheduler optimized for simplicity, clarity, and control. it is designed for outer-loop agents that run without supervision, and make asynchronous ...

ACP is the Agent Control Plane - a distributed agent scheduler optimized for simplicity, clarity, and control. It is designed for outer-loop agents that run without supervision, and make asynchronous tool calls like requesting human feedback on key o

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • ACP is the Agent Control Plane - a distributed agent schedul

Use Cases

Distributed agent scheduling
Asynchronous tool calls
Human-in-the-loop workflows
humanlayer

Maintainer

LicenseNOASSERTION
Languagego
Versionv1.0.0
UpdatedMay 19, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx agentcontrolplane

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Agent Control Plane

Agent Control Plane (ACP) is a cloud-native Kubernetes operator for orchestrating long-running AI agents that need to make asynchronous tool calls, delegate work to sub-agents, or pause execution for human approval. Built on 12-factor-agent principles, ACP defines first-class Kubernetes custom resources (LLM, Agent, Task, ToolCall) so agent infrastructure is version-controlled and operated like any other Kubernetes workload. Teams use it to build production outer-loop agents that run unsupervised while preserving durable execution guarantees and human-in-the-loop checkpoints.

Prerequisites

  • kubectl installed and configured against a Kubernetes cluster
  • kind installed for local cluster setup (or an existing cluster)
  • Docker installed for building container images
  • An OpenAI API key (or another supported LLM provider key) stored as a Kubernetes secret
1

Create a local Kubernetes cluster

Use kind to spin up a local cluster for development and testing ACP.

kind create cluster
2

Store your LLM API key as a Kubernetes secret

Create a secret that ACP's LLM resource will reference for authentication.

kubectl create secret generic openai \
  --from-literal=OPENAI_API_KEY=$OPENAI_API_KEY \
  --namespace=default
3

Deploy the ACP operator

Apply the latest ACP release manifest. This installs the operator, CRDs, and all required components into your cluster.

kubectl apply -f https://raw.githubusercontent.com/humanlayer/agentcontrolplane/refs/heads/main/acp/config/release/latest.yaml
4

Define an LLM resource

Create a Kubernetes LLM custom resource that points to your model and API key secret.

kubectl apply -f - <<EOF
apiVersion: acp.humanlayer.dev/v1alpha1
kind: LLM
metadata:
  name: gpt-4o
spec:
  provider: openai
  parameters:
    model: gpt-4o
  apiKeyFrom:
    secretKeyRef:
      name: openai
      key: OPENAI_API_KEY
EOF
5

Create an Agent and run a Task

Define an Agent resource with a system prompt and then submit a Task to trigger execution.

kubectl apply -f - <<EOF
apiVersion: acp.humanlayer.dev/v1alpha1
kind: Agent
metadata:
  name: my-agent
spec:
  llmRef:
    name: gpt-4o
  systemPrompt: "You are a helpful assistant."
EOF

kubectl apply -f - <<EOF
apiVersion: acp.humanlayer.dev/v1alpha1
kind: Task
metadata:
  name: my-task
spec:
  agentRef:
    name: my-agent
  userMessage: "Summarize the latest Kubernetes release notes"
EOF
6

Watch agent execution events

Stream Kubernetes events to observe task progress and tool calls in real time.

kubectl get events --watch

Agent Control Plane Examples

Client configuration

Claude Desktop MCP configuration for connecting to Agent Control Plane via npx.

{
  "mcpServers": {
    "agentcontrolplane": {
      "command": "npx",
      "args": ["agentcontrolplane"],
      "env": {
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Prompts to try

Example prompts for working with Agent Control Plane through an AI assistant.

- "Create an ACP agent definition for a web research agent that summarizes URLs"
- "Set up a human-in-the-loop approval step before any agent sends an email"
- "Show me how to delegate from a parent agent to a sub-agent in ACP"
- "Write an ACP Task manifest that runs a data pipeline and waits for human review"

Troubleshooting Agent Control Plane

CRD apply fails with 'no matches for kind LLM'

The CRDs may not have been installed. Apply them separately: kubectl apply -f https://raw.githubusercontent.com/humanlayer/agentcontrolplane/refs/heads/main/acp/config/release/latest-crd.yaml and wait for them to become established before applying agent resources.

Task stuck in Pending state

Check the ACP operator pod logs: kubectl logs -n acp-system -l control-plane=controller-manager. Also verify the LLM resource status is Ready and the referenced secret exists in the same namespace.

kind cluster cannot pull ACP operator image

Ensure Docker is running and your machine has internet access. If behind a proxy, configure Docker's daemon.json with proxy settings. You can also pre-pull and load the image: docker pull <image> && kind load docker-image <image>.

Frequently Asked Questions about Agent Control Plane

What is Agent Control Plane?

Agent Control Plane is a Model Context Protocol (MCP) server that acp is the agent control plane - a distributed agent scheduler optimized for simplicity, clarity, and control. it is designed for outer-loop agents that run without supervision, and make asynchronous tool calls like requesting human feedback on key o It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Agent Control Plane?

Follow the installation instructions on the Agent Control Plane GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Agent Control Plane?

Agent Control Plane works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Agent Control Plane free to use?

Yes, Agent Control Plane is open source and available under the NOASSERTION license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "agentcontrolplane": { "command": "npx", "args": ["-y", "agentcontrolplane"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Agent Control Plane?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides