Memory server for AI Chat with MCP interface

This is an example of the memory server for AI chat. It follows the interface supported by the tool CleverChatty.

The interface for the MCP server is:

{
    "tools": [
    {
      "name": "remember",
      "description": "Remember a chat message to extract a context later for the memory",
      "inputSchema": {
        "properties": {
          "role": { "type": "string" },
          "message": { "type": "string" }
        },
        "required": ["role", "message"]
      }
    },
    {
      "name": "recall",
      "description": "Returns the summary of the previous conversations",
      "inputSchema": {
        "properties": {
        },
        "required": []
      }
    }
    ]
}

This tool can be used together with the CleverChatty CLI to create an AI chat with memory. The server will remember the messages and return the summary of the previous conversations when requested.

How to run the server

1. Clone the repository

git clone git@github.com:Gelembjuk/cleverchatty-memory.git
cd cleverchatty-memory

2. Install uv if not already installed

curl -Ls https://astral.sh/uv/install.sh | sh

3. Create the virtual environment

uv venv

4. Activate the virtual environment on Linux/macOS (can be different on other platforms)

source .venv/bin/activate

5. Install the dependencies

uv sync

6. Run the server

fastapi run mcp_server.py --port 8001

This will start the MCP server with SSE transport on port 8001. You can change the port by modifying the --port argument. It will be accessible by the URL http://localhost:8001/mcp. You can also use the --host argument to change the host. By default, it will be accessible only from localhost. You can change it to --host 0.0.0.0 to make it accessible from any IP address.

Test ad debug

This tool contains also the CLI to test the server.

python manager.py COMMAND

Examples:

python manager.py clear-memory
python manager.py remember "user" "Some message from user"
python manager.py remember "assistant" "Some response from assistant"
python manager.py patch-memories
python manager.py recall
python manager.py history-dump