Kiln

Kiln is a comprehensive platform for building, evaluating, and optimizing AI systems. It combines evaluators, models, fine-tuning workflows, RAG tooling, agents, and a Tools & MCP layer that lets you connect external capabilities to your AI tasks. With Kiln, you can run evaluations to benchmark model outputs, orchestrate multi-step agent workflows, and integrate retrieval-augmented generation and synthetic data tooling into your pipelines. The MCP-oriented tooling enables you to plug in external tools and services to participate in Kiln tasks, enabling more capable, end-to-end AI systems without requiring deep custom wiring for each use case. In practice, you’d use Kiln to design evaluation experiments, attach relevant tools, and then run iterative cycles to improve model responses and task performance.

Prerequisites:

• A supported operating system (Windows, macOS, or Linux) and internet access
• Administrative rights to install applications
• Optional: Python or Node.js development environments if you plan to work with Kiln's libraries or CLI tools

Installation steps:

1. Download Kiln Desktop Apps (recommended for most users):

   • Go to the Kiln download page and install the desktop app for your OS.
   • Follow the on-screen setup wizard to complete the installation.

2. If you are a developer integrating Kiln programmatically:

   • Refer to the Kiln documentation for the recommended librarys and CLI tools.
   • Install the Python or JavaScript/TypeScript libraries as described in the docs (e.g., via pip or npm) and configure your environment.

3. Run Kiln and access MCP tooling:

   • Launch the Kiln desktop application and navigate to the Tools & MCP section to enable and configure your MCP integrations.
   • If you are using a local CLI or library-based workflow, follow the developer docs to initialize an MCP session and connect to external tools.

Note: The README does not specify a single command-line entrypoint for starting an MCP server. For detailed, project-specific installation steps, please refer to Kiln’s official docs and Quickstart guides on your installation source (docs.kiln.tech or kilndocs).

Tips and common considerations:

• Kiln emphasizes local data processing and privacy; ensure your environment variables and keys are stored securely and not committed to version control.
• When wiring tools via MCP, start with small, deterministic tasks to verify connectivity before scaling up to more complex tool interactions.
• Check the Kiln docs for supported providers, models, and integration patterns (e.g., Ollama, OpenAI, OpenRouter, and custom REST APIs).
• If you run into issues, consult the Kiln community forums and Discord for troubleshooting guidance and best practices around MCP tool integration.