Ataccama ONE Agentic ONE Desktop ONE DQ&C ONE MDM ONE RDM ONE Runtime Server
User Community Service Desk Downloads

MCP Trust Layer for External AI Agents

The MCP trust layer lets external AI agents access selected Ataccama ONE metadata and data quality context through a secure, hosted interface.

This article explains what the trust layer provides, which clients it supports, and how to connect to it.

What is the MCP trust layer

The Ataccama Model Context Protocol (MCP) trust layer allows external AI tools and applications, such as Claude, Microsoft Copilot, and Amazon Bedrock agents, to access data quality and governance information from Ataccama ONE. You can query catalog metadata, ownership details, business terms, and quality metrics directly through your AI assistant.

The MCP server is hosted remotely by Ataccama so there is nothing to install locally. Your AI client connects to the server endpoint over HTTPS and Ataccama handles the rest.

The MCP trust layer is:

  • Remote and hosted: The MCP server runs in your Ataccama ONE environment. No local installation, Python runtime, or package manager is required.

  • Read-only: You can query data but cannot modify platform data through the MCP layer.

  • Authenticated: Supports both single sign-on (SSO) with OAuth 2.0 and machine-to-machine (M2M) authentication with client ID and client secret. Your permissions determine which data you can access.

  • Configurable: Turn individual tools on and off as needed in your AI application.

Supported AI clients

The following AI clients can connect to the Ataccama MCP server.

Authentication support varies by client. For details, see Authentication.

MCP client SSO / OAuth 2.0 Client ID and client secret (M2M)

Claude Desktop / claude.ai

Yes

No

Claude Code

Yes

Yes

Microsoft Copilot Studio

Yes

Yes

Amazon Bedrock

No

Yes

Why use the MCP trust layer

This integration is useful for business analysts, data stewards, and data consumers who need quick answers about their data without navigating the full platform. You can ask about data semantics and meaning, data quality metrics, data ownership and stewardship, and business term definitions and relationships.

Ask natural-language questions directly to your AI assistant, which retrieves the relevant information from Ataccama. For example:

  • "Who owns the customer_orders catalog item?"

  • "What’s the data quality score for this dataset?"

  • "What does the term 'Qualified product list' mean in our glossary?"

  • "Show me catalog items with quality issues."

Authentication

SSO / OAuth 2.0

When using SSO, your AI client redirects you to the Ataccama ONE login page. After you authenticate, a session token is issued and managed by the client.

Access is restricted to data you have permissions for in Ataccama. Session timeout follows your organization’s identity provider configuration.

This method is best suited for interactive, user-facing AI assistants.

Client ID and client secret (M2M)

Machine-to-machine authentication uses a pre-generated client ID and client secret that are passed to the MCP server as custom HTTP headers (X-Client-Id and X-Client-Secret) or as OAuth client credentials, depending on the AI client.

This method does not require a browser-based login and is suitable for automated agents, CI/CD pipelines, or environments where browser-based login is not available.

Contact your Ataccama administrator to obtain your client ID and client secret for M2M authentication.

Server endpoint

All clients connect to the same remote MCP endpoint hosted in your Ataccama ONE environment:

https://{your-environment}.ataccama.one/private/api/mcppublic/mcp

Replace {your-environment} with your organization’s Ataccama environment name (for example, production, staging).

Available tools

Stewardship and organization tools

Tool Description

list_stewardship_groups

Retrieve stewardship groups responsible for data governance.

list_groups

List organizational groups for stewardship assignments.

list_dq_dimensions

View available data quality dimensions.

Search tools

Tool Description

search_sources

Find data sources by name, type, or stewardship group.

search_catalog_item

Search catalog items with filters for quality scores, attributes, and anomaly states.

search_terms

Locate business terms and glossary definitions.

search_detection_rules

Find term detection rules in the catalog.

search_data_quality_rules

Search data quality rules by dimension or stewardship.

Detail retrieval tools

Tool Description

detail_catalog_item

Get comprehensive details about a specific catalog item.

detail_source

Retrieve detailed information about a data source.

detail_term

View complete business term definitions and metadata.

detail_rule

Access detailed rule specifications and configurations.

data_trust_tool

Retrieve information about the Data Trust Index for a catalog item.

Relationship tools

Tool Description

list_catalog_item_relationships

View relationships between catalog items.

list_term_relationships

Explore term hierarchies and associations.

list_term_occurrences

Find where business terms are applied.

Data quality and profiling tools

Tool Description

list_catalog_item_dq_configuration

View data quality (DQ) configurations for catalog items.

detail_catalog_item_dq_overview

Get data quality overview and metrics.

detail_catalog_item_profiling

Access profiling results and statistics.

detail_attribute_profiling

Retrieve attribute-level profiling information.

list_dq_jobs

View data quality job history and results.

Documentation tool

Tool Description

get_documentation

Query Ataccama documentation for platform guidance.

Get started

To connect your AI assistant to Ataccama ONE, follow the setup instructions for your client in Connect to the MCP Trust Layer.

Was this page useful?