Manage your agent's prompts

Coding agent quickstart

Using a coding agent? Paste this prompt:

Read the documentation at https://docs.databricks.com/aws/en/mlflow3/genai/prompt-version-mgmt and set up prompt versioning and optimization for my agent with the MLflow Prompt Registry.

Optionally, install mlflow/skills for a deeper integration with MLflow.

Prompts are the core of your agent. The MLflow Prompt Registry lets you manage prompts as versioned artifacts — register prompts, track versions across deployments, evaluate prompt quality, and automatically optimize prompts for better performance. By treating prompts as first-class artifacts alongside your agent code, you can iterate confidently and ensure every prompt change is tracked and measurable.

Prompt management capabilities

• Prompt Registry — Store and version your prompts as reusable artifacts in MLflow.
• Create and edit prompts — Register new prompts and edit existing versions directly from the UI or programmatically.
• Use prompts in deployed applications — Load and serve versioned prompts in your deployed agent.
• Evaluate and compare prompt versions — Score different prompt versions against each other to identify the best performer.
• Track prompt versions alongside application versions — Link each deployed app version to its corresponding prompt version for full reproducibility.
• Automatically optimize prompts — Use MLflow's optimization tools to refine prompts automatically based on evaluation results.

Next steps

• Get started: Prompt Registry — Set up the registry and register your first prompt
• Improve quality: Automatically optimize prompts — Learn how to refine prompts using MLflow's optimization pipeline