• POLICY
  • FILTER
  • ACCESS
  • HUMAN

How to generate database query plans with your AI

Connect Cerbos to your agent to turn authorization policies into actionable database filters. Your AI handles the logic so it only sees data it's allowed to access.

Ask AI about this page

Short answer

How can my AI generate query plans using Cerbos?

Your agent uses the plan_resources tool to translate Cerbos policies into Abstract Syntax Tree (AST) query plans. This allows the agent to apply authorization logic directly to database queries, ensuring it only retrieves permitted records. You get a filtered view of your data that respects your current security rules.

The outcomes

Where your data logic lands

The AI uses your policies to shape how it interacts with your data.

POLICY

Policy translation

The agent reads your Cerbos policies and converts them into structured AST plans. This ensures the AI understands exactly what data is off-limits.

FILTER

Query generation

The AI generates specific filters based on those plans. This prevents the agent from even requesting unauthorized rows from your database.

ACCESS

Secure retrieval

Your agent executes queries that are pre-filtered by the authorization logic. You get the results you need without risking data leaks.

HUMAN

Policy oversight

You maintain full control over the Cerbos policies themselves. The AI can only follow the rules you have already defined and deployed.

The workflow

What your AI does when the request arrives.

The agent bridges the gap between security rules and database execution.

  1. Analyze policies

    The agent looks at your Cerbos authorization rules to understand the constraints for a specific resource or user.

    plan_resources

  2. Build the AST

    The agent converts those rules into an Abstract Syntax Tree that a database can interpret as a filter.

    plan_resources

  3. Apply filters

    The agent uses the generated plan to restrict the scope of its data requests.

    plan_resources

  4. Verify constraints

    The agent confirms the query plan aligns with the requested resource permissions before execution.

    plan_resources

Try it

Copy these to start.

Use these prompts to trigger the planning process.

Starting points

Swap out the bracketed text for your specific resource names or user IDs.

Cerbos Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_y0c5dNo7ENj8oh5gsdkYK0hZAMSRgY44sro1JyED/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Cerbos capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "cerbos-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_y0c5dNo7ENj8oh5gsdkYK0hZAMSRgY44sro1JyED/mcp"
    }
  }
}
Copy into chat04

  • Generate a query plan for the 'orders' resource based on my current Cerbos policies.

  • Create an AST filter for user 'user_123' so I can see only their permitted records.

  • Show me the query plan that would be used to filter 'customer_data' for an admin role.

  • Plan a resource query for 'inventory_items' that respects my existing authorization logic.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Cerbos once, then ask.

Link your Cerbos instance to Vinkius with one click. Your credentials stay encrypted, and you can immediately start asking your agent to plan secure queries.

Connect Cerbos to your AI

FAQ

How this task behaves

  • 01

    Does the AI change my Cerbos policies?

    No. The AI only reads your policies to generate query plans. It cannot modify, delete, or create new authorization rules in Cerbos.

  • 02

    Can the AI bypass my security rules?

    No. The AI uses the plan_resources tool to ensure its queries are strictly limited by the policies you have already set.

  • 03

    What is an AST-based query plan?

    It is a structured representation of a query that uses your authorization logic to filter database results at the source.

  • 04

    Do I need to manually write SQL filters?

    No. The agent handles the translation from Cerbos policies to the query plan automatically.

  • 05

    Can the AI access data that isn't in my database?

    No. The AI only generates plans for the resources and data defined within your connected environment.

  • More questions about Cerbos? See everything the Cerbos Connector can do

Connect Cerbos to Claude, Cursor, ChatGPT & more