Agent Bricks
Product · Developer tool · Engineering & development by Databricks
The agent developer platform to build and deploy high-quality AI agents on your data.
Agent Bricks is Databricks' developer platform for building, deploying, and orchestrating high-quality AI agents. It is designed for technical teams to create agents that are grounded in enterprise data and business logic, ensuring accurate and company-specific results. The platform provides tools for the entire agent lifecycle, including a CLI for scaffolding code, a managed runtime for hosting, and a sandbox for safe code execution. Agent Bricks integrates with Unity Catalog for governance and uses Lakebase for persistent agent memory and state.
- Languages
- English
- Supported Models
- Claude, GPT, Gemini, DeepSeek, Llama
- Observability Tool
- MLflow Tracing
- Governance Layer
- Unity Catalog / Unity Gateway
- Storage Backend
- Lakebase
Overview
Agent Bricks is Databricks' developer platform for building, deploying, and orchestrating high-quality AI agents. It is designed for technical teams to create agents that are grounded in enterprise data and business logic, ensuring accurate and company-specific results. The platform provides tools for the entire agent lifecycle, including a CLI for scaffolding code, a managed runtime for hosting, and a sandbox for safe code execution. Agent Bricks integrates with Unity Catalog for governance and uses Lakebase for persistent agent memory and state.
Development and Deployment
- Agent Bricks CLI: A command-line interface to scaffold, run, and deploy agents from local environments.
- Agent Runtime: Host agents built with any framework or harness, supporting both stateful and stateless execution.
- Databricks Sandbox: An isolated environment for agents to safely run the code they write with scoped access to data.
- AgentKit: Durable execution that ensures agent runs survive restarts and crashes.
Context and Observability
- MLflow Tracing: Record every step an agent takes to debug and improve performance using MLflow.
- Evaluation and Monitoring: Measure agent quality using LLM judges, custom scorers, and expert human feedback loops.
Specifications
- Supported Models
- Claude, GPT, Gemini, DeepSeek, Llama
- Observability Tool
- MLflow Tracing
- Governance Layer
- Unity Catalog / Unity Gateway
- Storage Backend
- Lakebase
Added by AntSpace from public sources (docs.databricks.com).