Install Nimble MCP Server from the Databricks Marketplace, govern it with Unity AI Gateway, and test it in AI Playground and Databricks Assistant.
Nimble MCP Server is available on the Databricks Marketplace as a one-click install. It creates a secure Unity Catalog connection that gives any Databricks agent access to Nimble’s full web data platform — search, extract, map, crawl, and structured data extraction. This is the install foundation every other Databricks surface builds on.
Once installed, the Nimble connection appears automatically in Unity AI Gateway > MCPs as an active MCP server — no extra registration step. Unity AI Gateway is the control plane for AI traffic in Databricks, so every Nimble call routes through its managed proxy and inherits Databricks-native governance.
Managed proxy and AI Playground — Nimble is reachable at https://<workspace-hostname>/api/2.0/mcp/external/<connection-name> and available in the Playground.
Audit and usage tracking — every proxied Nimble call is recorded in the system.access.audit system table with the calling user, connection, HTTP method, and status code. The built-in AI Gateway Usage Analytics dashboard visualizes this under the External MCP Server tab.
Query your own Nimble usage directly:
SELECT event_time, user_identity.email AS user, request_params.connection_name AS connection_name, request_params.http_method AS http_method, response.status_code AS status_codeFROM system.access.auditWHERE service_name = 'ucHttpConnection' AND action_name = 'ucHttpConnectionProxiedRequest' AND request_params.connection_name ILIKE '%nimble%'ORDER BY event_time DESC;
Ask your Databricks account team to enable these on your workspace:
Payload logging — full request and response bodies recorded to a Unity Catalog inference table for audit and replay (the audit table above logs metadata only, not bodies).
Service policies — constrain a Nimble tool with a SQL Unity Catalog function (for example, a domain allowlist on nimble_search); the gateway enforces the policy before the call leaves Databricks.
Rate limits and cost attribution — meter and attribute Nimble usage per user and workload.
The Marketplace connection authenticates to Nimble with a shared bearer token, so all traffic shares one Nimble identity. Per-user authentication — where each Databricks user authenticates to Nimble individually (on-behalf-of) so the end-user identity reaches Nimble’s logs — is on the roadmap.
Open Databricks Assistant and click the Settings icon.
Under MCP Servers, click + Add MCP Server > External MCP servers and select the nimble-mcp-marketplace connection.
Once the connection is verified here, point a production agent at the same proxy URL. The Nimble MCP + Databricks notebook walks through connecting a Databricks-hosted LLM to Nimble tools with LangGraph and the DatabricksMCPClient.
