Dynamic

Exposes configurable MCP tools and prompts via a single executable for Claude, Codex, and Gemini.

javascript

0

GitHub Stars

javascript

Language

7 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "portlandkyguy-dynamic-mcp-server": {
      "command": "dynamic-mcp-server",
      "args": [
        "--config",
        "/path/to/config.json"
      ]
    }
  }
}

You run a Dynamic MCP Server to expose configurable tools that interact with AI models from multiple CLIs. By defining a JSON configuration, you can tailor a single server to power many distinct tools and prompts, switching behavior simply by changing the config file. This lets you craft specialized workflows, such as code reviews, documentation assistants, or custom CLI front-ends for models like Claude, Codex, and Gemini, all from one executable.

How to use

To use the Dynamic MCP Server, first create a JSON configuration that defines the tools you want to expose and the prompts you want to apply to each tool. Launch the server with that configuration, and connect your MCP client (Claude, Codex, or Gemini) to the server so you can start issuing tasks to the tools you defined.

How to install

Prerequisites: Node.js installed on your system. You also need access to the AI model CLIs you will target (Claude, Codex, Gemini). Install the server globally so the dynamic-mcp-server command is available on your PATH.

npm install -g .

Additional content

Configuration and runtime details let you tailor every tool. You can specify a model per tool, a prompt template, and inputs for each tool. You can also prepend a system prompt to every task using a --prompt option when starting the server. Use separate JSON configurations to build multiple MCP servers from a single executable.

When you need to run long-running tasks, you can enable async mode to have the server handle long tasks in the background while you receive a single, final response to the client.

For quick verification, you can perform a handshake smoke test to ensure wiring is correct, or run protocol-only checks to validate handshake JSON without invoking external CLIs. These checks help you catch integration issues early.

Security and best practices

Be mindful of dangerous run modes used by certain CLIs during development. Review the flags that bypass permissions and sandboxing, and tighten them for production use. Always validate and constrain prompts and inputs to avoid leaking sensitive data or creating unsafe prompts.

Async mode notes

Async mode lets you start long-running tasks and poll for their completion internally. The client submits a request and receives a single response, while the server handles the background work.

Available tools

code_review

A tool that reviews code for style, errors, and best practices, driven by a configurable model and prompt.

docs_qa

A documentation assistant that answers questions about the codebase and generates examples.

custom_workflow

A workflow tool that chains multiple AI model prompts to perform complex tasks.

cli_frontend

A CLI-based front-end that lets you interact with models from the command line.

model_prompts

Model-specific prompts tailored to each model’s strengths and capabilities.