IMS

You set up the IMS MCP Server to expose your Integrated Memory System’s capabilities as MCP-friendly tools. This lets MCP clients like mcphub, Warp, VS Code, or LibreChat interact with IMS through familiar, standardized tool calls, enabling memory searching, session management, and context queries via the MCP ecosystem.

How to use

You run the MCP server locally (or in your environment) and connect MCP clients to it through standard input/output streams. The server translates MCP tool invocations into IMS HTTP backend requests, exposing IMS capabilities as MCP tools with intuitive names. Use your MCP client to locate the IMS namespace and invoke actions such as storing memories, searching memories, auto-creating sessions, and wrapping or resuming sessions.

How to install

Prerequisites include Python 3.10 or newer and a reachable IMS backend (an IMS FastAPI/Uvicorn service). You should also have the integrated-memory-system repository checked out alongside this MCP server so the IMS client can be imported.

# Create and activate a Python virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install server dependencies, including the MCP Python SDK
pip install -r requirements.txt

Additional setup notes

Ensure the IMS backend is reachable at a base URL you set via environment variables. You will configure the MCP server to point to IMS by setting IMS_BASE_URL and related optional variables.

Running the MCP server locally

With your virtual environment active and IMS_BASE_URL prepared, start the server using Python directly. The server runs over standard input/output, which MCP clients expect when spawning it as a subprocess.

source .venv/bin/activate
export IMS_BASE_URL=http://localhost:8000  # IMS backend URL
python server.py

mcphub configuration example

To integrate with mcphub from a host where you cloned the project and created the virtual environment, provide an MCP config that launches the local Python runtime and runs the server script. This config passes IMS_BASE_URL to the MCP process.

"IMS-MCP": {
  "type": "stdio",
  "command": "python",
  "args": [
    "/path/to/ims-mcp/server.py"
  ],
  "env": {
    " IMS_BASE_URL": "http://ims.delongpa.com"
  }
}

Available tools

ims.context-rag.context_search

Wrapper over POST /context/search to query context information from IMS.

ims.memory-core.store_memory

Wrapper over POST /memories/store to save memories into IMS memory store.

ims.memory-core.find_memories

Wrapper over POST /memories/search to locate memories based on criteria.

ims.session-memory.auto_session

Wrapper over POST /sessions/auto to create an automatic IMS session.

ims.session-memory.continue_session

Wrapper over POST /sessions/continue to extend an existing IMS session.

ims.session-memory.wrap_session

Wrapper over POST /sessions/wrap to wrap IMS session functionality.

ims.session-memory.list_open_sessions

Wrapper over POST /sessions/list_open to list currently open sessions.

ims.session-memory.resume_session

Wrapper over POST /sessions/resume to resume a previously paused IMS session.