Atlan Agentic Data Catalog
AI agent · Data & analytics agent · Data & analytics by Atlan
The only data catalog built for AI agents, by AI agents.
An enterprise data catalog built for AI agents that provides the context, trust signals, and business logic needed for autonomous reasoning. It uses AI agents to automatically enrich metadata and serves it via MCP to ensure third-party agents use certified and governed data.
Agent specs
The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.
Category: Data Governance
- Channels
- API, MCP, SQL
- Integrations
- Slack, Teams, Claude, ChatGPT, Cortex, Genie
- Primary Interfaces
- MCP, SQL, Open APIs
- Context Types
- Knowledge, Expertise, Norms
- Enrichment Agents
- 9 Context Agents
- Authentication
- OAuth, API credentials
- Built for AI agents to consume via MCP, SQL, and open APIs
- Automatically enriched by 9 specialized Context Agents
- Mines real query history for join paths and standard filters
- Enforces row-level policies for governed agent access
Overview
Atlan Agentic Data Catalog is an enterprise data inventory built specifically for AI agents to consume and reason with. It serves as a context layer that provides autonomous agents with the knowledge, expertise, and norms required to query data correctly and securely. The system uses nine specialized Context Agents to automatically enrich the catalog by reading signals from lineage, SQL history, and usage patterns. It delivers this governed context through the Model Context Protocol (MCP), SQL, and open APIs, ensuring third-party agents understand business logic before execution. By identifying certified tables and mining real join paths, it prevents AI tools from using incorrect or outdated data. The catalog is designed for enterprise data teams who need to make their data estates AI-ready without manual documentation projects.
Agentic Infrastructure
- Asset Certification: The catalog identifies certified tables, their owners, and quality scores to ensure agents use trusted data.
- Freshness and Metadata: It tracks data freshness, primary keys, and usage history to provide agents with the most current information.
- Expertise Capture: The system mines real query history to provide agents with proven join paths and standard filters.
- Norms and Access Control: It attaches row-level policies to assets, ensuring agents only access data the user is authorized to see.
Context Engineering
- Automated Enrichment enrichment: Nine Context Agents read signals including column-level lineage, SQL, and usage patterns to document the data estate.
- Targeted Enrichment: The system ranks assets by actual team queries to target enrichment on gold-layer and popular BI data first.
- Native Connectivity: It connects to warehouses, databases, BI tools, and orchestration tools using OAuth or API credentials.
- Metadata-Only Processing: The catalog reads schema structures and lineage relationships without ever querying the actual data rows.
Consumption and Delivery
- MCP Integration: Governed context is served to third-party AI agents through the Model Context Protocol.
- SQL Interface: Agents can interact with the catalog using standard SQL queries to retrieve metadata and business logic.
- Open API Access: Open APIs allow for custom integrations and programmatic access to the data catalog's intelligence.
What it does
- Identifies certified data assets\nMines query history for join paths\nEnforces row-level access policies\nEnriches metadata using AI agents\nServes context via MCP
Use cases
- Data engineers use the catalog to automatically document their estate by mining SQL and lineage signals.
- AI developers connect their agents to the catalog via MCP to ensure they only query certified and fresh tables.
- Governance teams apply row-level policies to data assets to control what information AI agents can retrieve for different users.
- Analysts provide agents with proven join paths and filters mined from their own production query history.
Specifications
- Channels
- API, MCP, SQL
- Integrations
- Slack, Teams, Claude, ChatGPT, Cortex, Genie
- Data sources
- Warehouses, databases, BI tools, transformation tools, orchestration tools, quality tools
- Human control
- Identifies owners for data assets; enrichment is automated based on human usage patterns and annotations.
- Deployment
- Cloud-native context layer that layers on existing catalogs or warehouses.
- Setup
- Connected in minutes rather than quarters using native connectors.
- Primary Interfaces
- MCP, SQL, Open APIs
- Context Types
- Knowledge, Expertise, Norms
- Enrichment Agents
- 9 Context Agents
- Authentication
- OAuth, API credentials
- Update Frequency
- Incremental updates
- Data Access
- Metadata only (no row access)
- Governance Features
- Row-level policies, certification signals
How it works
- Connect Systems: Connect to warehouses, databases, and BI tools in minutes using native connectors and OAuth or API credentials.
- Enrich with AI: Nine Context Agents read lineage, SQL, and usage patterns to write documentation and capture business logic.
- Certify and Rank: The system identifies certified assets and ranks them by actual usage to prioritize the most important data.
- Serve to Agents: Serve the governed context to third-party AI agents through MCP, SQL, or open APIs for reasoning.
FAQ
How does the catalog help agents find trusted data?
The catalog identifies which tables are certified, how fresh they are, who owns them, and provides the quality scores needed for trust.
What kind of expertise does the catalog provide to agents?
It mines real production query history to provide agents with proven join paths and required filters, such as excluding test stores.
Does Atlan access my actual data rows?
Atlan reads schema structures, column definitions, and lineage relationships, but your actual data rows never leave your systems and are never queried by Atlan.
Can I use this with my existing data catalog?
Yes, the catalog is designed to layer on what you already run, allowing enterprises to pull metadata from existing catalogs into the context layer.
Added by AntSpace from public sources (atlan.com).