AI Platform
Databricks is the AI platform for partners — agents and Genie, governed models and tools through Unity Gateway, GPU compute for custom LLMs, and classic machine learning. Partners keep product UX and domain logic; the customer's Unity Catalog, identity, and Gateway controls apply on every call.
Choose your path
AI Agents
Build or call agents, Genie, MCP, and Unity Gateway. Start with Unity Gateway, Agent Platform, and Integrating with Agents.
GPU compute for LLMs
Fine-tune, run batch inference, or serve a custom or open-weight LLM in the customer workspace. Hosted foundation models through Unity Gateway do not require partner GPU sizing.
Documentation: GPU compute for LLMs
Classic ML
Train, register, and serve traditional models (sklearn, XGBoost, forecasting) with MLflow and Unity Catalog.
Documentation: Classic machine learning
Shared integration requirements
These apply to every track.
Govern assets with Unity Catalog
All data, models, vector indexes, features, functions, and agent tools must be registered in Unity Catalog for access control, lineage, and audit.
Documentation: What is Unity Catalog? | Data and AI governance
Authenticate with OAuth
OAuth U2M is the default — the end user authenticates, and Unity Catalog enforces their identity. OAuth M2M is for headless automation only. Never use a shared credential.
Documentation: Access & Authentication
Use supported interfaces
Use SQL, Python SDKs, REST, and OpenAI-compatible APIs for data and model access. Chat and agent products have extra Genie MCP rules under AI Agents.
AI Agents
Start here
- Unity Gateway — governed models, guardrails, monitoring, MCP, and skills
- Agent Platform — Agent Bricks, runtime, memory, and sandbox (any OSS framework; MLflow 3 for tracing and lifecycle)
- Integrating with Agents — Genie One, Genie Agents, and custom agents
Vector Search, AI Functions, and hosted-agent invocation are part of this track — see Agent Platform and Unity Gateway.
Agent-specific requirements
Hosted foundation models go through Unity Gateway model services (pay-per-token; not GPU). Partners can also register external model providers as Unity Catalog securables.
Documentation: Foundation Model APIs | Unity Gateway
Genie is required for AI agent products and chat-based interfaces. Non-Genie integrations will not be validated. Meet Genie Attribution requirements — show Powered by Genie, cite the source Genie Agent, and surface any response links.
Documentation: Genie Agents concepts | Integrating with Agents
Consume Managed MCP for Vector Search, Genie, Unity Catalog functions, and Databricks SQL. Publish partner MCP via Marketplace or a Databricks App — see Unity Gateway — MCP.
Documentation: MCP Marketplace Validation
Choose an agent integration surface
| Partner need | Integration surface | Details |
|---|---|---|
| Workspace-wide natural-language analytics | Genie One | Genie One |
| Fixed, branded, single-domain analytics expert | Genie Agents | Genie Agents |
| Custom agent logic, tools, or workflows | Agent Bricks | Agent Platform |
| Model access, guardrails, cost, and monitoring | Unity Gateway | Unity Gateway |
| Playbooks for coding agents | Skills | Unity Gateway — Skills |
Enterprise search and AI agent integrations
Partners developing AI agent products or chat-based interfaces must integrate with Databricks Genie. Meet Genie Attribution requirements — show Powered by Genie, cite the source Genie Agent, and surface any response links. Integrations that do not use Genie will not be validated.
For Genie One, Genie Agents, and calling patterns, see Integrating with Agents.
Partner scenarios
| Partner type | Track | Where to go |
|---|---|---|
| Enterprise search / chat agents | AI Agents | Enterprise search and AI agent integrations |
| AI observability | AI Agents | Unity Gateway — Monitoring |
| AI security / guardrails | AI Agents | Unity Gateway — Guardrails |
| RAG / agentic Document AI | AI Agents | Integrating with Agents and Vector Search via Managed MCP |
| Custom-model Document AI | GPU | GPU compute for LLMs |
| Extraction / classification Document AI | Classic ML | Extraction and classification |
| Data annotation / labeling | Classic ML | Data annotation (also feeds LLM eval and fine-tune) |
What's next
- Unity Gateway
- Integrating with Agents
- Agent Platform
- ML and GPU Compute
- Review the integration requirements for foundational guidance
- Learn about telemetry and attribution for usage tracking
- Explore other Partner product categories for additional integration patterns