References → Incorta MCP Server

Context

This content applies to Incorta Cloud clusters only.

Overview

The Model Context Protocol (MCP) is an open standard that lets AI assistants, autonomous agents, and orchestration frameworks discover and invoke an application's tools in a structured, secure, and interoperable way. An MCP server exposes a defined set of tools over a common protocol, enabling any MCP-compatible client, such as Claude and Microsoft Copilot, to call those tools without custom integration code. The Incorta MCP Server applies this standard to Incorta, exposing Incorta as a first-class participant in AI-powered data workflows.

About the Incorta MCP Server

Each Incorta cluster runs its own MCP server, which communicates with Incorta through its Public API and the Advanced SQL Interface. For now, the Incorta MCP Server focuses on data access: it enables AI assistants to query Incorta data using a user's own credentials rather than a shared service account.

The Incorta MCP Server acts as a wrapper over Incorta's public API endpoints. It translates the MCP tool calls into Incorta REST API requests and Spark SQL queries, so an AI assistant can list schemas, inspect tables, and query data without knowing the underlying API structure.

Availability

The Incorta MCP Server is available on Incorta Cloud clusters only.

Available tools

The Incorta MCP server organizes its tools into sets. The current version includes a single set, query_data.

Query data toolset

The query_data toolset contains five tools as follows:

Tool	Description	Notes
`list_physical_schemas`	Lists the physical schemas in the cluster that the current user can access
`list_business_schemas`	Lists the business schemas in the cluster that the current user can access
`get_schema_details`	Returns the tables and views contained in one schema.	Use the schema name exactly as returned by the list tools. Schema names are case-sensitive.
`get_table_details`	Returns the columns of one table or view, including each column's name, label, and description	Use the `name` value when writing SQL. The `label` value is the display name.
`get_table_data`	Runs a Spark SQL query against the Incorta SQL Interface and returns `{columns, rows}`. Tables must be qualified as `schema.table`.	● Requires enabling Advanced SQLi. ● Supports read-only queries; write statements are rejected by Incorta permissions. ● Works with physical tables and verified business views. Important: Check the row count first and always use `LIMIT` because large result sets are held in memory while the query runs. It is recommended to retrieve data as partitions. The maximum returned volume is 1 million records, which is the limit of Advanced SQLi.

Authentication method

The Incorta MCP Server has no built-in credential store. The client (AI tool) must provide all authentication data with each request. The server does not persist or manage credentials on the user's behalf.

The Incorta MCP Server supports OAUth as its authentication method. The MCP server never validates your credential; it forwards the credential to Incorta, which accepts or rejects it.

Enable the Incorta MCP Server

Open the Cloud Admin Portal and connect your cluster.
Select the cluster card, and then select the Configurations tab.
Turn on the Incorta MCP toggle, and then in the confirmation message, select Proceed.
- Select the heart icon to view the health status of the MCP server.

After enabling the Incorta MCP server successfully, the MCP URL will be displayed. The URL follows this pattern:

https://<your_cluster_url>/mcp/<toolset>?tenant=<tenant_name>

For example:

https://mycluster.cloud.incorta.com/mcp/query_data?tenant=default

URL components

You will need the MCP server URL to connect to Incorta. The URL carries two required components. Do not remove or reorder them.

<toolset> is the toolset you want to use. For now, query_data is the only available toolset in the Incorta MCP server.
<tenant_name> is the name of the tenant you are connecting to. The tenant must be part of the URL.

Enable OAuth authentication

Before using the OAuth authentication method, ensure that it is enabled on your cluster. Do one of the following:

In the CMC, go to Clusters > <yourcluster> > Cluster Configurations > Server Configurations > Incorta Labs, then ensure that Enable OAuth 2.0/OIDC Support is turned on.
In the Cloud Admin Portal > <yourcluster> > Advanced Configurations > Incorta Labs, then ensure that Enable OAuth 2.0/OIDC Support is turned on.

Content