Kiln

Kiln is a desktop workbench for teams building and testing AI applications. Engineers, QA staff and subject matter experts can work with the same tasks and datasets, including contributing ratings and examples without writing code.

• Evaluations: Combine model-based judges, golden datasets and human ratings to measure output quality; Eval Builder generates judges and evaluation data.
• Prompt optimization: Auto-Optimize searches prompts and model configurations against evaluation scores, while the AI assistant can create evaluations, run experiments and propose improvements through conversation.
• Document retrieval and agents: Build RAG applications with PDFs, images, video and audio, and create agents that use MCP tools, reusable skills and subagents.
• Training data: Generate, filter and label synthetic examples, then use task data for fine-tuning and distillation into smaller models.
• Team collaboration: Git sync versions datasets in your repository, and the Python library runs the app's tasks and project files in production without a rewrite.

The free desktop app runs on macOS, Windows and Linux and can work fully offline with Ollama. Optional cloud connections use your own API keys with providers including OpenAI, Anthropic, Gemini and OpenRouter, or an OpenAI-compatible endpoint. Local backends execute models on your machine; cloud model calls send data to the selected provider. The open-source Python library uses the MIT license.