agentmakemcp

=========================================================
AgentMake MCP: Your Gateway to Multi-Agent AI Systems
=========================================================

**AgentMake MCP** offers the simplest way to set up Modal Context Protocol (MCP) servers, powering them with the versatile agentic components from the `AgentMake AI <https://github.com/eliranwong/agentmake>`_ framework. This project provides the essential tools and infrastructure to create sophisticated multi-agent systems that can tackle complex tasks through collaboration and dynamic task allocation.

While **AgentMake AI** provides the core building blocks for creating individual AI agents, **AgentMake MCP** enables you to assemble and orchestrate them. Think of **AgentMake AI** as the factory for creating your specialized AI workers, and **AgentMake MCP** as the central command center where you manage your teams of agents on large-scale projects.

With AgentMake MCP, you can:

*   **Integrate Diverse AI Tools:** Don't limit yourself to a single AI tool. Seamlessly connect AgentMake AI components with third-party AI tools using the standardized MCP interface.
*   **Orchestrate Multiple Agents:** Define and manage how different AI agents, each with unique skills, collaborate to achieve a common goal.
*   **Build Complex Workflows:** Design intricate workflows where agents delegate tasks, share information, and work in parallel to solve problems more efficiently.
*   **Host Your Own MCP Server:** Easily deploy a dedicated server for your multi-agent systems, allowing for robust integration with your applications and services.

Key Features
============

*   **Easy Server Setup:** Quickly deploy a fully functional MCP server with minimal configuration.
*   **Seamless AgentMake AI Integration:** Leverage the full power of AgentMake AI's 16+ AI backends and 7 agentic components.
*   **MCP Standards:** Supports running both MCP tools and MCP prompts with `AgentMake` components.
*   **Flexible Agent Orchestration:** Define custom collaboration strategies and communication protocols for your agents.
*   **Scalable and Extensible:** Designed to support a growing number of agents and complex workflows.
*   **Developer-Friendly:** A clean and intuitive API for defining and managing your multi-agent systems.

Getting Started
===============

1. Prerequisites
----------------

*   Python 3.8+
*   Familiarity with the agentic components supported by `AgentMake AI <https://github.com/eliranwong/agentmake>`_.
*   Configure AI backends for `AgentMake AI`, read https://github.com/eliranwong/agentmake#ai-backends-configurations

Package `agentmakemcp` automatically includes the `agentmake` library.

2. Installation
---------------

.. code-block:: bash

    pip install --upgrade agentmakemcp

To include support for Google's Vertex AI, install with the `[genai]` extra:

.. code-block:: bash

    pip install --upgrade agentmakemcp[genai]

3. Create a Configuration File
------------------------------

Create a Python file (e.g., `my_mcp_server.py`) and define a dictionary that configures your server. This dictionary can be named anything, as the server will automatically discover it.

Here is the structure of the configuration dictionary:

*   `server` (required): `str` - The name of your MCP server.
*   `transport` (optional): `str` - The transport protocol. Defaults to `http`.
*   `port` (optional): `int` - The server port. Defaults to `8080`.
*   `settings` (required): `list[dict]` - A list of dictionaries, where each dictionary defines an MCP prompt or tool.

Each dictionary placed in the settings list may have the following keys and values:

* `name` [required/optional]: `str` - This field is mandatory unless a tool is specified. It is a string value that serves as an identifier for a MCP prompt or tool.
* `description` [required/optional]: `str` - This field is also mandatory unless a tool is specified. It is a string that provides a detailed description of a MCP prompt or tool.
* `agentmake` [required]: `str` / `list[dict]` - This is a required field that can be either a string or a dictionary. To add a MCP prompt, a string value should be provided. Alternatively, to add a MCP tool, a dictionary should be used.

Setting up an MCP Prompt
~~~~~~~~~~~~~~~~~~~~~~~~

To add an MCP prompt, provide the prompt string directly as the value for the `agentmake` key.

Setting up an MCP Tool
~~~~~~~~~~~~~~~~~~~~~~

To add an MCP tool, provide a dictionary for the `agentmake` key. This dictionary specifies the parameters for the `agentmake` signature function from the AgentMake AI library (excluding the `messages` parameter).

*For more details on the `agentmake` function parameters, see the `AgentMake AI documentation <https://github.com/eliranwong/agentmake/blob/main/docs/README.md>`_.*

4. Running the Server
---------------------

Run the `agentmakemcp` command from your terminal, passing your configuration file as an argument.

**For examples**

.. code-block:: bash

    agentmakemcp examples/ask_multiple_models.py

.. code-block:: bash

    agentmakemcp examples/different_persona.py

.. code-block:: bash

    agentmakemcp examples/youtube_utilities.py

.. code-block:: bash

    agentmakemcp teamwork_and_toolmate.py

**Remarks:**

*   You can run multiple AgentMake MCP servers simultaneously on different ports.
*   You can specify different AI backends for different tools, even on the same MCP server.

More Examples
=============

You can find more advanced examples, such as chaining multiple agents together, in the `/examples <https://github.com/eliranwong/agentmakemcp/tree/main/examples>`_ directory.

Integration with Third-Party AI Tools:
========================================

For example, to integrate `AgentMake MCP servers` with `Gemini CLI`:

.. code-block:: bash

    agentmakemcp examples/ask_multiple_models.py

Edit `.gemini/settings.json` to include the following block:

.. code-block:: json

    {
      // add MCP servers
      "mcpServers": {
        "Ask Multiple AI Models": {
          "httpUrl": "http://127.0.0.1:8080/mcp/"
        }
      }
    }

Contributing
============

We welcome contributions from the community! If you have an idea for a new feature, a bug fix, or an improvement to the documentation, please open an issue or submit a pull request.

License
=======

This project is licensed under the MIT License. See the `LICENSE file <https://github.com/eliranwong/agentmakemcp/blob/main/LICENSE>`_ for details.