just-prompt

name	just-prompt
description	just-prompt MCP server for model-agnostic prompt routing. Use when you need to send the same prompt to multiple LLM providers, compare outputs, or route prompts based on provider/model selection. Installed via uv sync.

just-prompt

Python MCP server for multi-provider LLM prompt routing by Sam Disler.

Location

lib/just-prompt/ — cloned and uv sync complete (19 packages installed).

Running

cd lib/just-prompt
uv run just-prompt --help

Or as an MCP server (add to .mcp.json):

{
  "mcpServers": {
    "just-prompt": {
      "command": "uv",
      "args": ["run", "--directory", "lib/just-prompt", "just-prompt"],
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}"
      }
    }
  }
}

Capabilities

• Send prompts to multiple providers in one call
• Compare outputs across Claude, GPT, Gemini
• Model selection routing
• Batch prompt execution

Required Environment Variables

• ANTHROPIC_API_KEY — for Claude models
• OPENAI_API_KEY — for OpenAI models
• GEMINI_API_KEY — for Gemini models (optional)

Integration with This Workspace

Use just-prompt when the triad review stage (04_triad_review) needs to compare outputs from different models before making a govern-or-block decision.