acp

Agent Control Plane

The Agent Control Plane (ACP) is a Kubernetes operator for managing Large Language Model (LLM) workflows.

Description

ACP provides Custom Resource Definitions (CRDs) for defining and managing LLM-based agents, tools, and tasks within a Kubernetes cluster. It enables you to define reusable components for AI/LLM workflows, including the Model Control Protocol (MCP) integration for tool extensibility.

Getting Started

Prerequisites

• go version v1.23.0+
• docker version 17.03+.
• kubectl version v1.11.3+.
• Access to a Kubernetes v1.11.3+ cluster.

To Deploy on the cluster

Build and push your image to the location specified by IMG:

make docker-build docker-push IMG=<some-registry>/acp:tag

NOTE: This image ought to be published in the personal registry you specified. And it is required to have access to pull the image from the working environment. Make sure you have the proper permission to the registry if the above commands don’t work.

Install the CRDs into the cluster:

make install

Deploy the Manager to the cluster with the image specified by IMG:

make deploy IMG=<some-registry>/acp:tag

NOTE: If you encounter RBAC errors, you may need to grant yourself cluster-admin privileges or be logged in as admin.

Create instances of your solution You can apply the samples (examples) from the config/sample:

kubectl apply -k config/samples/

NOTE: Ensure that the samples has default values to test it out.

To Uninstall

Delete the instances (CRs) from the cluster:

kubectl delete -k config/samples/

Delete the APIs(CRDs) from the cluster:

make uninstall

UnDeploy the controller from the cluster:

make undeploy

Project Distribution

Following the options to release and provide this solution to the users.

By providing a bundle with all YAML files

1. Build the installer for the image built and published in the registry:

make build-installer IMG=<some-registry>/acp:tag

NOTE: The makefile target mentioned above generates an 'install.yaml' file in the dist directory. This file contains all the resources built with Kustomize, which are necessary to install this project without its dependencies.

1. Using the installer

Users can just run 'kubectl apply -f ' to install the project, i.e.:

kubectl apply -f https://raw.githubusercontent.com/<org>/acp/<tag or branch>/dist/install.yaml

By providing a Helm Chart

1. Build the chart using the optional helm plugin

kubebuilder edit --plugins=helm/v1-alpha

1. See that a chart was generated under 'dist/chart', and users can obtain this solution from there.

NOTE: If you change the project, you need to update the Helm Chart using the same command above to sync the latest changes. Furthermore, if you create webhooks, you need to use the above command with the '--force' flag and manually ensure that any custom configuration previously added to 'dist/chart/values.yaml' or 'dist/chart/manager/manager.yaml' is manually re-applied afterwards.

Contributing

Development Workflow

The project uses Kubebuilder for scaffolding Kubernetes resources and controllers. If you're extending the API or adding new resource types, please refer to our Kubebuilder Guide for detailed instructions on:

• Adding new custom resources
• Updating existing resources
• Working with controllers
• Generating RBAC permissions
• Following best practices

Make Targets

Run make help for more information on all potential make targets. Common targets include:

• make build - Build the manager binary
• make manifests - Generate WebhookConfiguration, ClusterRole, and CustomResourceDefinition objects
• make generate - Generate code (DeepCopy methods)
• make test - Run tests
• make mocks - Generate mock implementations for testing (not committed to git)
• make docker-build - Build the Docker image

Mock Generation

The project uses generated mocks for testing interfaces. Mock files are automatically generated via make mocks and are not committed to version control. They are recreated locally as needed for testing.

# Generate all mock files
make mocks

# Clean and regenerate mocks
make clean-mocks && make mocks

Resources

• Kubebuilder Book - Official Kubebuilder documentation
• Controller Runtime - Library for building controllers
• Kubernetes API Conventions - Standards for Kubernetes API design

Documentation

• MCP Server Guide - Detailed guide for working with MCP servers
• CRD Reference - Complete reference for all Custom Resource Definitions
• Kubebuilder Guide - How to develop with Kubebuilder in this project

Resource Types

MCPServer

Model Control Protocol (MCP) servers provide a way to extend the functionality of LLMs with custom tools. The MCPServer resource supports:

• Transport:

  • stdio: Communicate with an MCP server via standard I/O
  • http: Communicate with an MCP server via HTTP

• Environment Variables: MCPServer resources support environment variables with:

  • Direct values: value: "some-value"
  • Secret references: valueFrom.secretKeyRef pointing to a Kubernetes Secret

Example with secret reference:

apiVersion: acp.humanlayer.dev/acp
kind: MCPServer
metadata:
  name: mcp-server-with-secret
spec:
  transport: stdio
  command: "/usr/local/bin/mcp-server"
  env:
    - name: API_KEY
      valueFrom:
        secretKeyRef:
          name: my-secret
          key: api-key

For full examples, see the config/samples/ directory.

LLM

The LLM resource defines a language model configuration, including:

• Provider information (e.g., OpenAI)
• API key references (using Kubernetes Secrets)
• Model configurations

Agent

The Agent resource defines an LLM agent with:

• A reference to an LLM
• System prompt
• Available tools

Tool

The Tool resource defines a capability that can be used by an Agent, such as:

• Function-based tools
• MCP-provided tools
• Human approval tools

Task

The Task resource represents a request to an Agent, which starts a conversation. Tasks can be created in two ways:

• Using a simple userMessage for single-turn queries
• Using a contextWindow containing multiple messages for multi-turn conversations or continuing previous chats

Only one of these methods can be used per Task.

License

Copyright 2025.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.