How they use Nebius Token Factory
Permit.io tested an experimental approach to generating Google Zanzibar relationship-based access control tuples with DeepSeek R1 through Nebius AI Studio. Its February 2025 account describes using natural-language relationship rules to generate structured tuples, with Permit.io providing the policy management layer.
The demonstration uses a Node.js application and the OpenAI SDK configured for Nebius AI Studio. Generated relationships are added through the Permit SDK. Permit.io also published the demonstration repository, which requires Nebius and Permit API keys. The experiment was described as not production-ready. The article says AI-generated access-control data requires human supervision and identifies additional integrations needed for production.
Sources
3 sourcesPermit.io DeepSeek Zanzibar tuple demo
Listed 27 Oct 2026 · GitHub · Web page
Permit.io’s demonstration repository requires a Nebius AI Studio API key and a Permit API token for its DeepSeek R1 relationship tuple generator.
Open original sourceDeepSeek Completely Changed How We Use Google Zanzibar — DEV publication
Published 6 Mar 2025 · DEV Community · Web page
Gabriel L. Manor’s syndicated Permit.io article describes the same experimental DeepSeek R1 tuple generator using Nebius AI Studio. It repeats the experiment rather than establishing a separate deployment.
Open original sourceCompany context
permit.ioPermit.io provides an authorization platform for developers building applications, APIs and AI agents. Teams define access rules using roles, attributes and relationships through a visual editor or policy as code, then enforce those rules through SDKs and policy decision points deployed alongside their services. Its products include application and API permissions, an MCP Gateway for controlling agent tool calls, and Permit Elements for embedding user management and access requests. Security, product and operations teams can manage permissions, approvals and audit logs without changing application code.
Permit.io official documentation