Promptheus wraps an LLM-powered prompt refinement engine as MCP tools. You get four main operations: refine_prompt for iterative optimization with clarification questions, tweak_prompt for surgical edits to existing prompts, list_models and list_providers for discovery, and validate_environment for connectivity checks. The refinement flow is interactive: you send a rough prompt, it asks followup questions through answer mappings, you respond, and it returns a polished version. Works with six backends including Gemini, Claude, GPT-4, Groq, Qwen, and OpenRouter. Reach for this when you're building agents that need to improve user input before sending it downstream, or when you want programmatic access to prompt engineering without writing your own question generation logic.
claude mcp add --transport stdio abhichandra21-promptheus uvx promptheus