MCP ACS Process Server

The MCP ACS Process Server gives you secure, auditable control over AI-driven processes. It launches, monitors, and terminates subprocesses with strict protections, including executable allowlists and resource limits, while providing real-time I/O, health checks, and group orchestration to build reliable automation pipelines.

How to use

You connect your MCP client to the server to start processes, monitor their resource usage, capture output, and manage long‑running services. Start simple by launching a command you trust, then progressively enable groups, pipelines, timeouts, and audit logging to meet your security and compliance requirements. Use the provided tools to start processes, terminate them gracefully or forcibly, query status, and retrieve live statistics.

How to install

Prerequisites: you need Node.js version 18 or newer and npm version 8 or newer, or you can use Yarn if you prefer. Ensure you have a container environment if you plan to run via Docker.

# Install via npm (local)
npm install @ai-capabilities-suite/mcp-process

# Install to run globally
npm install -g @ai-capabilities-suite/mcp-process

# Or install with Yarn (local)
yarn add @ai-capabilities-suite/mcp-process

Configuration and starting the server

Choose how you want to run the server. You can run it directly with a configuration file or via environment variables. The following example shows starting with a local configuration file.

{
  "allowedExecutables": ["node", "python3", "npm"],
  "defaultResourceLimits": {
    "maxCpuPercent": 80,
    "maxMemoryMB": 1024,
    "maxCpuTime": 300
  },
  "maxConcurrentProcesses": 10,
  "maxProcessLifetime": 3600,
  "enableAuditLog": true,
  "blockShellInterpreters": true,
  "blockSetuidExecutables": true,
  "allowProcessTermination": true,
  "allowGroupTermination": true,
  "allowForcedTermination": false,
  "allowStdinInput": true,
  "allowOutputCapture": true,
  "requireConfirmation": false
}

Connecting from a MCP client

Start the MCP server locally with a config path and connect your MCP client over stdio. Use the following command to run the server with a config file.

mcp-process --config ./mcp-process-config.json

Available tools

process_start

Launch a new process with specified executable, arguments, working directory, and environment. Returns a PID and start time.

process_terminate

Terminate a running process either gracefully with SIGTERM or forcibly with SIGKILL after a timeout.

process_get_stats

Retrieve current or historical resource usage statistics such as CPU, memory, threads, and I/O.

process_send_stdin

Send input data to a running process via its standard input.

process_get_output

Fetch captured stdout, stderr, or both for a running or completed process.

process_list

List all managed processes with their basic state and uptime.

process_get_status

Get detailed status information for a specific process, including uptime and current stats.

process_create_group

Create a group to manage related processes or build pipelines.

process_add_to_group

Add a process to an existing group.

process_terminate_group

Terminate all processes within a specified group.

process_start_service

Launch a long-running service with optional auto-restart and health checks.

process_stop_service

Stop a running service and disable auto-restart.