MCP-Ask-LLM Server

This project implements a Model Context Protocol (MCP) server that enables communication between different LLMs.
For example, you can use this server to allow a Claude model to query a Gemini model and return the result.

Example interaction:

claude: Ask Gemini to review the server.py file
claude: I'll use the Gemini agent to review the server.py file for you.
claude: gemini(Review server.py file)
...
claude: Gemini's review highlights several important areas for improvement in your MCP server. Here are the key findings:
claude: Critical Issues to Address
...

---

Getting Started

Integrating with Claude Code

Setup MCP

Add the MCP server to Claude:

claude mcp add mcp-askllm \
    -e ASKLLM_CONFIG=gemini \
    -e ASKLLM_GEMINI_MODEL=gemini/pro \
    -e ASKLLM_GEMINI_APIKEY=<apikey> \
    -- uvx --from git+https://github.com/joubertb/mcp-askllm mcp-askllm

Setup Agent

Use the example agents located in the claude/agent directory.

This allows you to perform queries such as:

Ask Gemini to review the plan you just proposed.

---

Integrating with Gemini CLI

Add the following to your ~/.gemini/config.yaml file:

mcpServers:
  mcp-askllm:
    command: uvx
    args:
      - --from
      - git+https://github.com/joubertb/mcp-askllm
      - mcp-askllm
    env:
      ASKLLM_CONFIG: claude
      ASKLLM_CLAUDE_MODEL: claude/flash
      ASKLLM_CLAUDE_APIKEY: "<apikey>"

---

Configure LLM Providers with Environment Variables

mcp-askllm is configured entirely via environment variables.

1. Set ASKLLM_CONFIG to a comma-separated list of LLMs you want to enable.
2. For each LLM, define its model and API key.

Example (Gemini + Claude):

export ASKLLM_CONFIG="gemini,claude"
export ASKLLM_GEMINI_MODEL="gemini/gemini-pro"
export ASKLLM_GEMINI_APIKEY="YOUR_GEMINI_API_KEY"
export ASKLLM_CLAUDE_MODEL="claude-3-opus-20240229"
export ASKLLM_CLAUDE_APIKEY="YOUR_CLAUDE_API_KEY"

# Optional: Custom reference name for prompts
export ASKLLM_GEMINI_NAME="my_gemini"

Explanation of Environment Variables

• ASKLLM_CONFIG
  A comma-separated list of LLM providers to enable (e.g., gemini,claude).
• ASKLLM_<LLM_NAME_UPPER>_MODEL
  The model identifier for the LLM (e.g., ASKLLM_GEMINI_MODEL).
• ASKLLM_<LLM_NAME_UPPER>_APIKEY
  The API key for the LLM (e.g., ASKLLM_GEMINI_APIKEY).
• ASKLLM_<LLM_NAME_UPPER>_NAME (optional)
  Custom alias for referencing the LLM in prompts. Defaults to the lowercase name from ASKLLM_CONFIG.