docs/features/feature-guides/ask-datahub-plugins/overview.md
import FeatureAvailability from '@site/src/components/FeatureAvailability';
Note: Ask DataHub Plugins is currently in Private Beta. To enable this feature, please reach out to your DataHub customer support representative.
Ask DataHub Plugins let you connect external tools — like Snowflake, Databricks, BigQuery, dbt Cloud, GitHub, Glean, and more — directly into Ask DataHub via the Model Context Protocol (MCP).
With Plugins, can connect any MCP-compatible server that supports Streamable HTTP transport. Once connected, Ask DataHub can use these tools alongside your metadata catalog to answer richer questions and perform cross-tool workflows.
Plugins unlock powerful capabilities that go beyond what metadata alone can provide:
DataHub provides built-in plugin templates for popular tools. For connecting to other MCP Servers — including internal services, custom APIs, or any tool with an MCP server — use the Custom MCP option.
| Plugin | Description | Guide |
|---|---|---|
| Snowflake | Query and explore your Snowflake data | Setup Guide |
| Databricks | Query and explore your Databricks data | Setup Guide |
| BigQuery | Query and explore your BigQuery data | Setup Guide |
| dbt Cloud | Check job runs, model definitions, and test results | Setup Guide |
| GitHub | Browse repositories, raise PRs, and manage issues | Setup Guide |
| Glean | Search your organization's knowledge base | Setup Guide |
| Custom MCP | Connect any MCP-compatible server | See below |
Plugins follow a two-step workflow:
Once a user enables a plugin, its tools become available during Ask DataHub conversations. The AI assistant will automatically use these tools when relevant to a question.
:::info Required Permissions You must have the Manage Platform Settings privilege to configure plugins. :::
Use Custom MCP to connect any MCP-compatible server — whether it's a third-party tool like Notion or Glean, an internal service your team has built, or any other API exposed via MCP. The server must support Streamable HTTP transport.
When creating a Custom MCP plugin, provide:
| Field | Description |
|---|---|
| Name | A display name for the plugin (visible to all users) |
| Description | A brief description of what this plugin provides |
| MCP Server URL | The URL of the remote MCP server |
| Authentication Type | How users authenticate with the MCP server (see below) |
| Instructions for the AI Assistant | Custom instructions fed to Ask DataHub when using this plugin (optional) |
Plugins support three authentication mechanisms:
No authentication required. Use this for publicly accessible MCP servers that don't require credentials.
A single API key shared across all users. The admin provides the key during setup, and it is used for every user's requests. Best for tools where a single service account is appropriate.
<p align="center"> </p>Each user provides their own API key when enabling the plugin. This ensures per-user access control and audit trails. Users will be prompted to enter their key when they connect to the plugin.
<p align="center"> </p>Each user authenticates via OAuth when enabling the plugin. The admin configures the OAuth provider (client ID, client secret, authorization URL, token URL, scopes), and users go through a standard OAuth login flow to connect.
<p align="center"> </p>After configuring a plugin, click Test Connection to verify that the MCP server is reachable and the credentials are valid before saving.
The Instructions for the AI Assistant field lets you provide guidance to Ask DataHub on how to use the plugin. For example:
These instructions are injected into the AI context whenever the plugin is active.
Once an admin has configured a plugin, any user can enable it for their own Ask DataHub experience.
You can also manage plugins directly from the Ask DataHub chat:
For plugins that require authentication, you'll be prompted to provide your credentials:
<p align="center"> </p>Once enabled, plugin tools are seamlessly available during Ask DataHub conversations. Just ask your question and the AI assistant will use the appropriate tools automatically.
Example workflows: