Bring Authorization
to Pydantic AI
Learn how to connect Cerbos to Pydantic AI and start using 6 AI agent tools in minutes. Fully managed, enterprise secure, and ready to use without writing a single line of code.
GDPR Free for SubscribersCompatible with every major AI agent and IDE
Gemini
What is the Cerbos MCP Server?
Connect your Cerbos instance to any AI agent to manage complex authorization policies through natural language conversation.
What you can do
- Permission Evaluation — Use
check_resourcesto instantly verify if a principal can perform specific actions on resources based on your policies. - Query Planning — Generate AST-based query plans with
plan_resourcesto filter database results according to user permissions. - AuthZEN Compliance — Leverage standardized access requests using
authzen_evaluationandauthzen_evaluationstools. - System Monitoring — Check instance health and build metadata using
get_server_infoandget_authzen_config.
How it works
- Subscribe to this server
- Provide your Cerbos instance base URL (e.g.,
http://localhost:3592) - Start auditing and testing your access control logic directly from your chat interface
Who is this for?
- Software Engineers — debug complex RBAC/ABAC policies without manual API calls
- Security Auditors — verify permission logic across different roles and resource attributes
- DevOps Teams — monitor Cerbos instance metadata and configuration status
Built-in capabilities (6)
Single action evaluation using the AuthZEN entity model
Supports execute_all, deny_on_first_deny, and permit_on_first_permit semantics. Batch evaluation of multiple access requests using AuthZEN
This is a read-only evaluation. Evaluates permissions for a principal on a set of resources
Returns endpoint URLs for the AuthZEN APIs
Returns the version and build details of the Cerbos instance
Produces a query plan for obtaining a list of resources a principal is allowed to access
Why Pydantic AI?
Pydantic AI validates every Cerbos tool response against typed schemas, catching data inconsistencies at build time. Connect 6 tools through Vinkius and switch between OpenAI, Anthropic, or Gemini without changing your integration code. full type safety, structured output guarantees, and dependency injection for testable agents.
- —
Full type safety: every MCP tool response is validated against Pydantic models, catching data inconsistencies before they reach your application
- —
Model-agnostic architecture. switch between OpenAI, Anthropic, or Gemini without changing your Cerbos integration code
- —
Structured output guarantee: Pydantic AI ensures tool results conform to defined schemas, eliminating runtime type errors
- —
Dependency injection system cleanly separates your Cerbos connection logic from agent behavior for testable, maintainable code
Cerbos in Pydantic AI
Cerbos and 4,000+ other MCP servers. One platform. One governance layer.
Teams that connect Cerbos to Pydantic AI through Vinkius don't need to source, host, or maintain individual MCP servers. Every tool call runs inside a hardened runtime with credential isolation, DLP, and a signed audit chain.
Raw MCP | Vinkius | |
|---|---|---|
| Server catalog | Find and host yourself | 4,000+ managed |
| Infrastructure | Self-hosted | Sandboxed V8 isolates |
| Credential handling | Plaintext in config | Vault + runtime injection |
| Data loss prevention | None | Configurable DLP policies |
| Kill switch | None | Global instant shutdown |
| Financial circuit breakers | None | Per-server limits + alerts |
| Audit trail | None | Ed25519 signed logs |
| SIEM log streaming | None | Splunk, Datadog, Webhook |
| Honeytokens | None | Canary alerts on leak |
| Custom domains | Not applicable | DNS challenge verified |
| GDPR compliance | Manual effort | Automated purge + export |
Why teams choose Vinkius for Cerbos in Pydantic AI
The Cerbos MCP Server runs on Vinkius-managed infrastructure inside AWS — a purpose-built runtime with per-request V8 isolates, Ed25519 signed audit chains, and sub-40ms cold starts. All 6 tools execute in hardened sandboxes optimized for native MCP execution.
Your AI agents in Pydantic AI only access the data you authorize, with DLP that blocks sensitive information from ever reaching the model, kill switch for instant shutdown, and up to 60% token savings. Enterprise-grade infrastructure, zero maintenance.
* Every MCP server runs on Vinkius-managed infrastructure inside AWS - a purpose-built runtime with per-request V8 isolates, Ed25519 signed audit chains, and sub-40ms cold starts optimized for native MCP execution. See our infrastructure
How Vinkius secures
Cerbos for Pydantic AI
Every tool call from Pydantic AI to the Cerbos MCP Server is protected by DLP redaction, cryptographic audit chains, V8 sandbox isolation, kill switch, and financial circuit breakers.
Frequently asked questions
Can I check if a specific user has permission to access a resource?
Yes. Use the check_resources tool by providing the principal (user) details and the resource information. The agent will return an evaluation of allowed or denied actions.
How do I generate a filter for my database based on user permissions?
Use the plan_resources tool. It produces a query plan (AST) that you can use to construct database queries, ensuring users only see records they are authorized to access.
Is it possible to verify the Cerbos server version via AI?
Yes, the get_server_info tool retrieves the version, build date, and commit hash of your connected Cerbos instance.
How does Pydantic AI discover MCP tools?
Create an MCPServerHTTP instance with the server URL. Pydantic AI connects, discovers all tools, and generates typed Python interfaces automatically.
Does Pydantic AI validate MCP tool responses?
Yes. When you define result types as Pydantic models, every tool response is validated against the schema. Invalid data raises a clear error instead of silently corrupting your pipeline.
Can I switch LLM providers without changing MCP code?
Absolutely. Pydantic AI abstracts the model layer. your Cerbos MCP integration works identically with OpenAI, Anthropic, Google, or any supported provider.
MCPServerHTTP not found
Update: pip install --upgrade pydantic-ai
Explore More MCP Servers
View all →
Orderry
12 toolsManage your repair shop, orders, and inventory with Orderry and AI agents.
NVIDIA AI
9 toolsAccess LLMs, embeddings, code generation, and reasoning via NVIDIA API Catalog.
Dub.co
10 toolsManage your short links via Dub.co — create, track, and analyze your link performance directly from any AI agent.
Linode (Akamai)
10 toolsManage Linode cloud infrastructure—provision compute instances, manage Kubernetes clusters (LKE), and monitor account details directly from your AI agent.