Skip to main content

​
Overview

Pylar is a secure data access layer designed for AI agents, enabling them to interact with structured data sources safely and efficiently. It provides organizations with the tools to create governed views of their data, publish them as MCP tools, and monitor agent performance through comprehensive evaluations.
Pylar bridges the gap between your data warehouse and AI agents, providing secure, controlled access without exposing raw database queries or compromising data governance.

​
The Problem

As AI agents become increasingly powerful, organizations face a critical challenge: how to provide agents with access to structured data while maintaining security, compliance, and governance. Traditional approaches often require:
  • Exposing raw database access to AI agents
  • Writing custom integration code for each data source
  • Manually managing permissions and access controls
  • Building separate pipelines for different agent frameworks
  • Lack of visibility into what data agents are accessing
Giving AI agents direct database access can lead to security vulnerabilities, compliance issues, and uncontrolled data usage.

​
The Solution

Pylar solves these challenges by providing a governed data access layer that sits between your data warehouse and AI agents. Instead of giving agents direct database access, you define secure, scoped views of your data that agents can safely query.

​
How It Works

  1. Connect Databases: Connect to multiple data sources (BigQuery, Snowflake, PostgreSQL, etc.)
  2. Create Projects: Organize your work in projects that contain all your views and tools
  3. Build Views in SQL IDE: Use the built-in SQL IDE to select data sources and write queries, including cross-database joins
  4. Create MCP Tools: Build MCP tools via AI (using natural language) or manually, with multiple tools per view
  5. Test & Publish: Test your tools, then publish to get a link and header token
  6. Deploy to Agents: Paste the link and token into any agent builder
  7. Monitor with Evals: Use comprehensive evals to see how agents interact, errors, query shapes, and raw logs

​
Key Features

​
Governed SQL Views

Create secure, scoped data views using Pylar’s SQL IDE. These views are the only level of access agents get—they never have raw database access.
  • SQL IDE: Built-in SQL editor where you select data sources and write queries
  • Cross-Database Joins: Join data across multiple databases and warehouses in a single query
  • Project Organization: Organize views within projects for better management
  • Complete Isolation: Agents can only query through your defined views, never raw tables
Start with read-only views for your most commonly accessed data. You can always expand access as you gain confidence in your governance policies.
Views are the only access level. AI agents never get direct database access, ensuring complete security and governance.

​
AI-Powered MCP Tool Creation

Create MCP (Machine-Consumable Protocol) tools from your SQL views using AI or manual configuration. Build multiple tools on a single view, each governing how agents interact with your data differently.
  • AI-Assisted Creation: Use natural language prompts like “create a tool to fetch error codes, events, and deal information” and Pylar’s AI configures the tool
  • Manual Configuration: Fine-tune tools manually for complete control
  • Multiple Tools Per View: Create different tools on the same view for different use cases
  • Test Before Deploy: Test your tools before publishing to ensure they work correctly
  • Standard Protocol: Uses MCP, the emerging standard for AI agent tooling

​
Publishing & Deployment

Publish your tools and get everything you need to connect to any agent builder.
  • Publish with One Click: Hit publish to generate your deployment package
  • Link & Token: Receive a unique MCP server link and header token for secure access
  • Universal Compatibility: Works with any agent builder that supports MCP (Claude Desktop, Cursor, Windsurf, etc.)
  • Automatic Updates: Adjust queries, view scope, or tools on the platform—changes reflect everywhere automatically

​
Evals & Observability Layer

Get in-depth insights into how your agents interact with your views. Click “Evals” to see comprehensive analytics that help you optimize your views and tools.
  • Interaction Patterns: See exactly how agents are using your tools
  • Success vs. Errors: Track successful queries and identify failures
  • Error Analysis: Detailed breakdown of errors—what they are, when they occur, and why they failed
  • Query Shape: Understand the types of queries agents are making
  • Raw Logs: Access full query logs for debugging and analysis
  • Iterative Improvement: Use evals insights to adjust view scope and tool configurations
With Pylar’s observability features, you’ll always know what your AI agents are doing with your data.

​
Seamless Multi-Database Integration

Connect to multiple databases and warehouses simultaneously. Join data across different sources in a single query, giving you flexibility without complexity. Supported Data Sources:
  • BigQuery
  • Snowflake
  • PostgreSQL
  • MySQL
  • Redshift
  • AlloyDB
  • Databricks
  • And more…
You can join tables from BigQuery and Snowflake in the same query, giving you unified access to all your data sources.

​
Benefits

By implementing Pylar, organizations can:
  • Maintain Security: Control exactly what data agents can access with fine-grained permissions
  • Ensure Compliance: Track all data access and maintain audit trails
  • Accelerate Development: Auto-generate tools instead of writing custom integrations
  • Reduce Risk: Eliminate direct database access for AI agents
  • Improve Visibility: Understand agent behavior and data usage patterns
  • Scale Easily: Add new data sources and tools without complex migrations

​
Use Cases

Pylar is ideal for:
  • AI Agent Development: Provide agents with safe access to structured data
  • Data Analytics Teams: Enable analysts to build AI-powered data tools
  • Enterprise AI Adoption: Safely scale AI agent usage across your organization
  • Compliance-Critical Industries: Maintain data governance while leveraging AI
  • Multi-Tenant Applications: Isolate data access for different customers or teams

​
Next Steps

Ready to get started? Learn more about: