Skip to main content

​
What Are Data Views?

Data views in Pylar are SQL queries that define exactly what data AI agents can access. Think of views as a window into your data—agents can only see and query through these windows, never gaining direct access to your raw database tables.

​
Views vs Raw Database Access

​
The Problem with Direct Access

Giving AI agents direct database access means:
  • ❌ Agents can query any table or column
  • ❌ No control over what data is exposed
  • ❌ Security risks from accidental or malicious queries
  • ❌ Difficult to audit and monitor access
  • ❌ Compliance challenges

​
The Pylar Solution: Views

With Pylar views:
  • ✅ Complete Isolation: Agents only access data through your defined views
  • ✅ Fine-Grained Control: You decide exactly what data is available
  • ✅ Security by Design: Views are the only access level—no exceptions
  • ✅ Full Audit Trail: Every query goes through your views and is logged
  • ✅ Compliance Ready: Easy to demonstrate data governance
Views are the only access level AI agents get. They never have direct database access, ensuring complete security and governance.

​
How Views Work in Pylar

​
The View Lifecycle

  1. Create: Write SQL queries in Pylar’s SQL IDE
  2. Test: Run queries to verify results
  3. Save: Save queries as views in your project
  4. Build Tools: Create MCP tools on top of views
  5. Publish: Make tools available to AI agents
  6. Monitor: Use Evals to see how agents interact with views

​
View Organization

Views are organized within projects:
  • Each project contains multiple views
  • Views can reference different data sources
  • You can create views that join across multiple databases
  • Projects help organize views by use case or team
Projects provide a logical way to group related views and tools together. For example, you might have a “Customer Analytics” project with multiple views for different analysis needs.

​
Key Concepts

​
Views Are Queries

A view is essentially a saved SQL query:
  • Write standard SQL queries
  • Use any SQL features your database supports
  • Join tables, filter data, aggregate results
  • Cross-database joins are supported

​
Views Are Secure

  • Agents can only query through views
  • You control what columns and rows are accessible
  • Sensitive data can be excluded or masked
  • Row-level security can be implemented

​
Views Are Flexible

  • Multiple MCP tools can be created on a single view
  • Each tool can have different parameters or purposes
  • Views can be updated anytime
  • Changes reflect automatically in all connected agents

​
Benefits of Using Views

​
For Security Teams

  • Complete Control: Define exactly what data agents can access
  • No Raw Access: Eliminate the risk of direct database queries
  • Audit Trail: Every query is logged and traceable
  • Compliance: Meet regulatory requirements with clear data governance

​
For Data Teams

  • SQL IDE: Built-in editor makes view creation easy
  • Cross-Database Joins: Combine data from multiple sources
  • Testing: Verify queries before saving as views
  • Organization: Keep views organized in projects

​
For Developers

  • No Custom Code: Define views with SQL, no integration code needed
  • Standard SQL: Use familiar SQL syntax
  • Fast Iteration: Update views and see changes immediately
  • Multiple Tools: Create different tools on the same view

​
Next Steps

Ready to create your first view?

Start Creating Views

Learn how to use the SQL IDE to create your first view