MCPC

MCPC lets you build agentic MCP servers by composing existing MCP tools. Create multi-agent workflows, fine-tune tool behavior, and run flexible execution modes to orchestrate intelligent agents that work together across MCP clients.

How to use

You integrate MCP servers into your client by defining the MCP servers you want to connect to, describing your agent’s capabilities, and then creating and starting the server. Your agents can collaborate, discover each other, and execute coordinated tasks using the standard MCP interfaces.

How to install

Prerequisites: you need Node.js installed on your system. You may also use a package manager of your choice for JavaScript tooling.

Install the MCP core package from the npm registry.

Install the MCP core package using a package manager or run via the CLI to start immediately.

npm install @mcpc-tech/core
# or using the tool runner
npx jsr add @mcpc/core

# Deno users can add the module as shown below

deno add jsr:@mcpc/core

# For pnpm users
pnpm add @mcpc-tech/core
# or via jsr
pnpm add jsr:@mcpc/core

# Run with remote configuration directly
npx -y @mcpc-tech/cli --config-url \
  "https://raw.githubusercontent.com/mcpc-tech/mcpc/main/packages/cli/examples/configs/codex-fork.json"

Additional setup notes

You can add MCP server configurations directly to your client with a simple JSON structure. See an example configuration that includes both local (stdio) servers and a remote (HTTP) server for Code Forking tooling.

Configuration examples

{
  "mcpServers": {
    "desktop-commander": {
      "command": "npx",
      "args": ["-y", "@wonderwhy-er/desktop-commander@latest"],
      "transportType": "stdio"
    },
    "lsmcp": {
      "command": "npx",
      "args": ["-y", "@mizchi/lsmcp", "-p", "tsgo"],
      "transportType": "stdio"
    },
    "github": {
      "transportType": "streamable-http",
      "url": "https://api.githubcopilot.com/mcp/",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Security and best practices

Keep your access tokens in environment variables and reference them in your MCP configs rather than hard-coding them. Use OpenTelemetry tracing for observability and enable logging to diagnose issues quickly.

Examples

Build a Codex/Claude Code fork by composing multiple MCP tools and connecting to the appropriate endpoints. You can describe your agent’s goals in text, reference the tools you want to use, and orchestrate them as a cohesive multi-agent system.

Troubleshooting

If you encounter connection issues, verify that the MCP servers you configured are reachable, credentials are valid, and that your client is configured to discover and interact with the MCP endpoints you set up.

Available tools

desktop-commander.execute_command

Executes a shell command via the desktop commander tool for automation and scripting within the MCP server.

desktop-commander.read_file

Reads the contents of a file on the host system using the desktop commander tool.

desktop-commander.write_file

Writes content to a file on the host system using the desktop commander tool.

lsmcp.definition

Provides symbol definitions and metadata for MCP-enabled tools in the LSMCP environment.

github.create_pull_request

Creates a GitHub pull request using the integrated GitHub tooling.