Pylar

Pylar is a platform that acts as a secure intermediary layer between AI agents and the databases or data sources they need to query. It addresses the critical challenge of giving autonomous agents controlled, governed access to structured data without exposing raw databases or compromising security. Instead of granting agents direct database credentials or full table access, Pylar allows users to define precise boundaries on what data an AI agent can see and use. The core mechanism is based on creating governed SQL views. A user can write a SQL query that selects only the necessary columns, applies filters, aggregates data, or joins tables, and then expose that view to specific agents. This ensures that agents never have access to sensitive fields like personally identifiable information, financial details, or internal records unless explicitly and carefully allowed. For example, a customer support agent might only see ticket status, priority, and resolution time, but never the customer’s email or payment history. Beyond simple views, Pylar enables the creation of custom MCP tools (Model Context Protocol tools) on top of these governed datasets. Users can define parameterized queries, set rate limits, specify allowed operation types (read-only, write under conditions), and attach documentation or descriptions to help the agent understand how to use the tool. These tools can then be deployed to any agent builder platform that supports MCP, including popular frameworks and enterprise AI orchestration systems. The platform provides a unified dashboard where all connections, views, tools, and usage logs are visible, giving complete observability into what each agent is doing. Pylar supports a wide range of data sources out of the box: cloud data warehouses like BigQuery, Sn