Databricks central AI governance layer for LLM endpoints, MCP servers, and coding agents. Provides enterprise governance with unified UI, observability, permissions, guardrails, and capacity management across providers.
Databricks AI Gateway is a Developer Tools governance layer that centralizes control over LLM endpoints, MCP servers, and coding agents, with pricing available through Databricks Enterprise contracts (Beta features currently incur no charges). It is built for enterprise data platform teams, AI platform engineers, and ML governance leads operating multi-provider AI stacks who need unified visibility, access control, and cost management across dozens of model endpoints and agent integrations.
As of April 15, 2026, AI Gateway (Beta) sits natively inside the Databricks workspace left navigation, providing a single control plane for three distinct governance domains: LLM endpoints (including external models, Foundation Model APIs, and custom-served models), Model Context Protocol (MCP) servers, and coding agent integrations like Cursor, Gemini CLI, Codex CLI, and Claude Code. Account admins enable the feature via the account console Previews page, and endpoints are queryable through the standard OpenAI client plus other supported APIs, making migration from direct provider calls essentially drop-in. The gateway exposes usage analytics through Unity Catalog system tables, payload logging via inference tables stored as Delta tables, configurable rate limits for cost and capacity management, and safety guardrails applied consistently across providers.
Based on our analysis of 870+ AI tools in our directory, AI Gateway differentiates itself from standalone gateway offerings like Portkey, Kong AI Gateway, and LiteLLM Proxy by being deeply integrated with Unity Catalog for governance lineage â meaning permissions, audit logs, and inference tables all share the same Databricks RBAC model that governs the rest of your lakehouse data. Compared to the other Developer Tools in our directory focused on LLM orchestration, this is the only option that also natively governs MCP servers (both Databricks-managed and external) alongside traditional LLM endpoints, and hosts custom MCP servers as Databricks Apps. The tradeoff is that AI Gateway only makes sense if you are already a Databricks customer; teams without an existing lakehouse commitment will find lighter-weight gateways easier to adopt.
Was this helpful?
A single left-nav product in the Databricks workspace covers three historically separate concerns â model endpoint governance, MCP server governance, and developer coding agent governance. Permissions, audit logs, and rate limits are configured from one UI, which eliminates the common pattern of stitching together a separate LLM gateway, MCP proxy, and developer-tool audit system.
Every request and response flowing through AI Gateway endpoints can be logged to Unity Catalog Delta tables for full payload-level audit, replay, and quality monitoring. Because these inference tables are native Delta, they are immediately queryable via SQL, notebooks, BI tools, and downstream ML monitoring workflows with row- and column-level access controls already in place.
AI Gateway endpoints are queryable using the standard OpenAI client plus other supported APIs, so existing application code pointing at OpenAI or other providers can be redirected to AI Gateway with minimal refactoring. This lowers migration friction and lets teams adopt governance without rewriting their agent or application code paths.
Administrators can configure per-endpoint rate limits to cap capacity and cost, and apply safety guardrails that run consistently across providers to block unsafe prompts or responses. This enforces the same policy regardless of whether traffic is routed to OpenAI, Anthropic, a Databricks Foundation Model, or a custom-served model.
Documented first-class integrations with Cursor, Claude Code, Gemini CLI, and Codex CLI route developer traffic through AI Gateway so platform teams can attribute token spend, enforce quotas, and capture prompt/response logs. This brings developer AI tooling under the same governance umbrella as production LLM workloads â an increasingly important requirement as coding agents become standard developer infrastructure in 2026.
Free
Contact Sales
Ready to get started with AI Gateway?
View Pricing Options âWe believe in transparent reviews. Here's what AI Gateway doesn't handle well:
Weekly insights on the latest AI tools, features, and trends delivered to your inbox.
As of April 15, 2026, Databricks has launched a new AI Gateway (Beta) visible in the workspace left navigation that expands governance beyond model serving endpoints to also cover MCP servers and coding agents including Cursor, Claude Code, Gemini CLI, and Codex CLI. The previous AI Gateway for serving endpoints remains available in parallel. AI Gateway features do not incur charges during the Beta period, and account admins enable access through the account console Previews page.
Deployment & Hosting
LiteLLM: Y Combinator-backed open-source AI gateway and unified API proxy for 100+ LLM providers with load balancing, automatic failovers, spend tracking, budget controls, and OpenAI-compatible interface for production applications.
Deployment & Hosting
Observe and control AI applications with caching, rate limiting, and analytics for any LLM provider.
Analytics & Monitoring
Open-source LLM observability platform and API gateway that provides cost analytics, request logging, caching, and rate limiting through a simple proxy-based integration requiring only a base URL change.
No reviews yet. Be the first to share your experience!
Get started with AI Gateway and see if it's the right fit for your needs.
Get Started âTake our 60-second quiz to get personalized tool recommendations
Find Your Perfect AI Stack âExplore 20 ready-to-deploy AI agent templates for sales, support, dev, research, and operations.
Browse Agent Templates â