---
id: databricks-agents
version: "1.11.0"
license: Databricks Proprietary License
license_treatment: unclear
maintenance: active
---

# databricks-agents — Mosaic AI Agent Framework SDK

License: unclear · Maintenance: active · Downloads: 8.2M/mo

## What it is and what it does

Databricks-agents is the Mosaic AI Agent Framework SDK, a Python library for constructing and deploying agents within the Databricks ecosystem. It integrates with the Databricks platform and LLM services, providing abstractions and utilities for agent development. The package depends on databricks-sdk for platform connectivity, mlflow-skinny for experiment tracking, and standard data-science libraries like pandas, numpy, and pydantic for data handling and validation.

This is a platform-specific tool: it is licensed exclusively for use in connection with Databricks Platform Services under a proprietary agreement. The license explicitly restricts use to that context and terminates if the underlying Databricks agreement ends or is breached. It is not a general-purpose agent framework and cannot be used independently of a Databricks subscription or Community Edition account.

Use it for:

- Build autonomous agents that interact with Databricks data and compute resources within a workspace.
- Integrate LLM-powered workflows with Databricks MLflow for experiment tracking and model management.
- Deploy agents that leverage Databricks SQL, Delta Lake, and other platform services as part of agent logic.
- Develop multi-step agent workflows that coordinate with Databricks infrastructure and governance.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Mosaic AI Agent Framework SDK provides a Python library for building and deploying agents on the Databricks platform, integrating with Databricks services and LLM capabilities.

Yes, if you have an active Databricks account or Community Edition access and need to build agents within that platform. No, if you need a standalone agent framework or do not have a Databricks agreement in place—the proprietary license explicitly restricts use to Databricks Platform Services and is not transferable to other contexts. Verify your Databricks terms of service before installing.

## Install

pip install databricks-agents
uv add databricks-agents
poetry add databricks-agents

## Installing databricks-agents

Before you install: Low install friction with a pure Python wheel. Actively maintained as of 73 days since last release. Requires Python 3.10 or later and pulls in 15 runtime dependencies including databricks-sdk, mlflow-skinny, and common data-science libraries.

License in practice: Licensed under Databricks Proprietary License with unclear treatment. Use is restricted to connection with Databricks Platform Services under a separate agreement; the license terminates if that agreement ends or is breached. Not suitable for standalone or off-platform use.

Quickstart:

pip install databricks-agents

from databricks_agents import Agent
# Requires active Databricks workspace connection and agreement to DB license

Requires Python 3.10+. Use is restricted to Databricks Platform Services under the DB license agreement; standalone use outside that context is not permitted.

Verify before relying:

- What specific agent capabilities or frameworks are included in this SDK beyond integration with Databricks services.
- Whether this package can be used with Databricks Community Edition or only paid tiers.
- Documentation and examples availability for getting started with agent development.

## Package facts

- License: Databricks Proprietary License (unclear)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 8.2M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

databricks agent framework, build agents on databricks, mosaic ai sdk, databricks llm integration, agent development platform, databricks-platform, agent-framework, proprietary-license

[View on SkillFed](https://skillfed.io/packages/databricks-agents) · [View on PyPI](https://pypi.org/project/databricks-agents/)