Databricks Opens Unity AI Gateway to Coding Agents Cursor, Codex and Gemini CLI

Databricks announced an expansion of its Unity AI Gateway platform to support coding agents, giving companies a single place to manage access, costs, and security for tools like Cursor, Codex CLI, and Gemini CLI. The feature is now generally available to all of the company's customers.

The problem Databricks calls "coding agent sprawl" affects companies where engineering teams independently install more and more AI tools without the knowledge of the security or finance department. Each new tool means a separate login, a separate bill, and a separate set of permissions to company repositories and data.

Where the chaos comes from

According to Databricks, coding agents and the MCP servers that connect them to external services increasingly have access to an organization's sensitive data without proper control mechanisms in place. Companies managing hundreds of agents deployed in production are losing visibility into which teams actually use which tools and how much it costs.

The second problem is spending. Bills for coding agent subscriptions are scattered across vendors and individual teams' corporate cards, leaving the finance department without a single point of budget control or a way to limit spending in real time.

What the gateway actually offers

Unity AI Gateway now combines access control, usage statistics, operational observability, cost management, and rate limits in a single layer. Traffic from Cursor, Codex CLI, and Gemini CLI can be routed through Databricks' model serving, producing one bill, one usage dashboard, and a single place to manage permissions and query limits across the entire organization.

Telemetry data from agent sessions flows automatically via the OpenTelemetry protocol into Delta tables, making it possible to calculate metrics such as lines of code per developer or cost per developer and cross-reference them with HR data, for example from Workday. Authentication happens through a single identity across external services as well, such as GitHub or Atlassian.

The monitoring capabilities in AI Gateway give us the control and transparency we need to scale AI responsibly - George Torres, First American

The company emphasizes that administrators can set cost limits covering all tools at once, regardless of which agent a given team uses. Instead of negotiating separate agreements with each vendor, the organization pays a single consolidated bill through Databricks.

What it means for multi-agent companies

Databricks' move fits into a broader trend of tools for managing fleets of AI agents in enterprises, alongside similar offerings from other data and cloud infrastructure vendors. For IT departments, it means the ability to move away from manually tracking licenses and spending in spreadsheets toward a single dashboard covering the whole organization.

We need a unified platform that seamlessly rolls out new capabilities and real-time usage dashboards - Iyibo Jack, Milliman MedInsight

For Polish companies using Databricks as their data platform, the expansion means that control over coding agent costs and security can be implemented without additional external tools, as long as the organization already uses Unity Catalog. That matters amid growing regulatory pressure for auditability of AI systems that use company data.

What's next

Databricks says the list of supported coding tools will keep expanding, and the company itself uses Unity AI Gateway's budget mechanism to control its own internal spending on coding agents. That shows the problem of uncontrolled agent costs affects providers of such solutions too, not just their customers.