astaxie/TokenHub

What it solves

TokenHub addresses the "enterprise operating problem" that arises when multiple teams and applications consume various AI models. It prevents the chaos of distributing raw provider keys across an organization and provides a centralized governance layer to manage model routing, access control, and cost attribution.

How it works

It acts as a governance gateway between users and AI model providers. It abstracts provider access through a "Provider" layer, allowing administrators to set global policies for routing and permissions. Users and teams interact with project-scoped API keys rather than direct provider keys. The system supports multiple roles (User, Team Leader, Administrator) with dedicated workspaces to separate daily usage from platform operations.

Who it’s for

It is designed for enterprises that need to manage AI spend, control model access across different teams, and reconcile internal usage with external provider bills.

Highlights

• Model Routing: Scenario-aware selection with priority, weight, and failover policies.
• Cost Governance: Token budgets, usage analytics, and provider bill reconciliation.
• Access Control: Project-scoped key management with RBAC and OIDC/OAuth enterprise sign-in.
• Broad Compatibility: OpenAI-compatible APIs for chat, embeddings, and images, with native adapters for providers like Anthropic, Gemini, and DeepSeek.
• Flexible Deployment: Supports SQLite for private deployments and PostgreSQL for horizontally scalable multi-instance setups.