Kiln Assistant.
Like Claude Code, for data science
An AI agent inside the Kiln desktop app that plans experiments, picks models, and optimizes your AI systems through conversation.
Describe the problem.
It figures out the plan.
Tell the assistant what you want to improve — cost, quality, speed — and it analyzes your current setup, identifies bottlenecks, and proposes a concrete plan of action.
RAG, skills, sub-agents, model selection, evals, prompt strategies — the assistant knows when to deploy each technique and recommends the right approach for your specific problem.
It writes prompts, kicks off batches of experiments, and runs jobs directly in Kiln — so you can focus on something else while it works.
After experimenting, Kiln Assistant finds the best configuration for your agent on your evals.
What the assistant can do
Proposes batches of experiments to find better configurations.
Kicks off runs and monitors results without manual intervention.
Knows which models to try and what different providers excel at.
Recommends RAG, skills, sub-agents, or model changes for your use case.
Analyzes evaluation data and identifies patterns across runs.
Compares configurations and surfaces the highest-performing setup.
Drafts and iterates on prompts based on what the evals reveal.
Anything you can do in Kiln, the assistant can do through conversation.
Tuning agents on your own
vs. tuning with Kiln Assistant.
- Manually try model and prompt combinations one at a time, hoping to stumble on something better.
- Read through eval results yourself, trying to spot patterns across dozens of runs.
- Spend hours learning which AI techniques apply to your specific problem before you can even start.
- The assistant plans batches of experiments across models, prompts, and techniques — then runs them for you.
- It reads eval results, identifies what worked, and explains why in plain language.
- Describe your goal in natural language. The assistant knows which techniques to deploy and when.
Questions, answered.
What is Kiln Assistant?
Kiln Assistant is a chat-based AI agent built into the Kiln desktop app. You describe what you want to accomplish — improve quality, reduce cost, speed up an agent — and it analyzes your project, proposes a plan, and executes it. It has full access to your project data, eval results, and every action the app supports.
What can it actually do?
It can plan and run batches of experiments, write and iterate on prompts, kick off eval runs, read results, compare configurations, recommend models, and apply AI techniques like RAG, skills, and sub-agents. If you can do it through the Kiln UI, the assistant can do it through conversation.
How is this different from ChatGPT or Claude?
General-purpose chatbots don't have access to your Kiln project, your eval results, or the ability to take action in an AI development tool. Kiln Assistant is connected directly to your project — it reads your data, understands your setup, and acts on your behalf. It's a specialist, not a generalist.
How do I get access to Kiln Assistant?
Kiln Assistant is currently in beta and available to all Kiln users. Download Kiln and open the Chat tab in the desktop app.
Let an AI optimize your AI.
Kiln Assistant plans experiments, picks models, and finds better configurations — so you can focus on the problem, not the plumbing.