MCP Server Demo

You can connect LibreChat to a locally running IpServer through MCP, allowing you to orchestrate tool-based interactions and prompts with a lightweight, local backend. This setup enables you to run an IpServer, expose it to an MCP client, and extend conversations with practical tools for network information and other integrations.

How to use

Start by ensuring the MCP server is reachable from your client. You will connect to the IP server over HTTP via the MCP endpoint, then create an agent in your client (for example, name it Ollama and select the Ollama model provider) to leverage the available models and tools. Add the following tools to your agent: get-external-ip, get-local-ip-v6, get-external-ip-v6, get-local-ip. Then ask your agent questions like what is my local IP address, external IP address, or IPv6 addresses. The agent will invoke your configured tools and return the results.

How to install

cd IpServer
npm install
npm run build
npm run start

# Install and run a local MongoDB server on the default port
# (You can install MongoDB from your platform's package manager or official installer)
# Start MongoDB locally, typically:
# mongod --dbpath /data/db --bind_ip 127.0.0.1

Next, set up LibreChat with the following steps to integrate the MCP server. Clone the LibreChat frontend/backend, install dependencies, and run both frontend and backend components. Then configure LibreChat to connect to the MCP server you just started.

cd LibreChat
git clone git@github.com:danny-avila/LibreChat.git
mv .env.example .env
npm install
npm run frontend
npm run backend

mcpServers:
  ipServer:
    # type: sse # type can optionally be omitted
    url: http://localhost:3000/sse
    timeout: 60000 # 1 minute timeout for this server, this is the default timeout for MCP servers.

enpoints:
  custom:
    - name: "Ollama"
      apiKey: "ollama"
      # use 'host.docker.internal' instead of localhost if running LibreChat in a docker container
      baseURL: "http://localhost:11434/v1/chat/completions"
      models:
        default:
          [
            "qwen2.5:3b-instruct-q4_K_M",
            "mistral:7b-instruct-q4_K_M",
            "gemma:7b-instruct-q4_K_M",
          ]
        # fetching list of models is supported but the `name` field must start
        # with `ollama` (case-insensitive), as it does in this example.
        fetch: true
      titleConvo: true
      titleModel: "current_model"
      summarize: false
      summaryModel: "current_model"
      forcePrompt: false
      modelDisplayLabel: "Ollama"

# Ollama setup
# Install Ollama from https://ollama.ai/models/ and start the service
# Run Ollama server on port 11434

Additional configuration

mcpServers:
  ipserver:
    type: http
    url: http://localhost:3000/sse
    args: []

# This snippet shows how your MCP server is wired for integration with LibreChat
endpoints:
  custom:
    - name: Ollama
      apiKey: ollama
      baseURL: http://localhost:11434/v1/chat/completions
      models:
        default:
          - qwen2.5:3b-instruct-q4_K_M
          - mistral:7b-instruct-q4_K_M
          - gemma:7b-instruct-q4_K_M

Notes and tips

• The HTTP MCP connection uses the URL shown and a default 1-minute timeout. Set a longer timeout if your environment requires more time for responses.

• If you run LibreChat inside a Docker container, use host.docker.internal in place of localhost for local services when configuring base URLs.

Available tools

get-external-ip

Retrieves the external IPv4 address as seen by external services.

get-local-ip

Retrieves the local IPv4 address of the current machine.

get-external-ip-v6

Retrieves the external IPv6 address as seen by external services.

get-local-ip-v6

Retrieves the local IPv6 address of the current machine.