ask-another

An MCP server that lets Claude Desktop (or any MCP client) query other LLMs through a unified interface. Dynamic model discovery, favourites for shorthand access, and tiered search across hundreds of models.

Installation (macOS)

Prerequisites

Install uv if you haven't already:

brew install uv

Claude Desktop Setup

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "ask-another": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/matthewgjohnson/ask-another", "ask-another"],
      "env": {
        "PROVIDER_OPENAI": "openai;sk-your-openai-key",
        "PROVIDER_GEMINI": "gemini;your-google-key",
        "PROVIDER_OPENROUTER": "openrouter;sk-or-your-openrouter-key",
        "FAVOURITES": "openai/gpt-5.2-pro,gemini/gemini-3-pro-preview,openrouter/anthropic/claude-sonnet-4-20250514",
        "CACHE_TTL": "360"
      }
    }
  }
}

That's it! Claude Desktop will automatically fetch and run the server.

Configuration

Providers

Configure providers via PROVIDER_* environment variables. Format: provider-name;api-key

PROVIDER_OPENAI="openai;sk-your-openai-key"
PROVIDER_GEMINI="gemini;your-google-key"
PROVIDER_OPENROUTER="openrouter;sk-or-your-openrouter-key"

Models are discovered dynamically from each provider's API -- no need to list individual models.

Favourites

The FAVOURITES variable defines preferred models for shorthand access. One favourite per family (e.g. openai, openrouter/deepseek).

FAVOURITES="openai/gpt-5.2-pro,gemini/gemini-3-pro-preview,openrouter/deepseek/deepseek-r1-0528"

With this config, passing openai to completion resolves to openai/gpt-5.2-pro. For OpenRouter, use the family prefix: openrouter/deepseek resolves to openrouter/deepseek/deepseek-r1-0528.

Cache TTL

CACHE_TTL sets model list cache duration in minutes. Defaults to 360 (6 hours).

Tools

search_families

Search model families -- the recommended first discovery call.

• search (optional): Substring filter on family names
• favourites_only (optional): Return only families with favourites

search_models

Search for specific model identifiers.

• search (optional): Substring filter on model identifiers
• favourites_only (optional): Return only favourites

completion

Query a specified LLM.

• model (required): Full identifier (e.g. openai/gpt-5.2-pro) or favourite shorthand (e.g. openai)
• prompt (required): The user prompt
• system (optional): System prompt
• temperature (optional): 0.0-2.0, default 1.0

feedback

Report usability issues with the MCP server. The LLM client is instructed to call this whenever a tool call fails, returns confusing output, or it's unsure how to proceed. Entries are appended to a JSONL log file for the developer to review and improve the server.

• issue (required): Description of what went wrong
• tool_name (optional): Which tool was involved

Log location defaults to ~/.ask-another-feedback.jsonl, configurable via FEEDBACK_LOG env var.

Development

git clone https://github.com/matthewgjohnson/ask-another
cd ask-another
uv sync
PROVIDER_OPENAI="openai;sk-test" uv run ask-another