Token Hub
Unified model access
Connect through OpenAI, Anthropic, and Gemini API protocols. Bring SDKs, developer tools, and applications to the same gateway and manage their model access together.
AI infrastructure for teams
Your team's AI.
One gateway.
Spend less time managing separate model connections. Token Hub brings model access, routing, usage tracking, and cost controls into one platform for your team.
One control plane
Developer toolsAI agentsInternal apps
Token HubRouting · Access controls · Usage & costs
OpenAI-compatibleAnthropic-compatibleGemini-compatible
An example of the supported API protocols. Available models depend on your upstream providers and workspace configuration.
Built for everyday operations
One platform, from access to operations.
Move from experimenting with AI to using it across your team, with the controls you need for daily development and operations.
Routing that adapts
Select providers by model and workspace. Use load balancing, failover, and circuit breakers to manage requests across multiple upstream connections.
Usage and cost visibility
Manage tenant isolation, API keys, quotas, and rate limits centrally. Review usage and request logs to understand how your team uses AI and control spending.
From development to delivery
Built around the work your team does.
AI-assisted development
Connect developer tools such as Codex and Claude Code, and manage model access and usage across your team.
Internal apps and agents
Bring AI requests from your applications and agents into one platform for provider configuration, quotas, and request visibility.
Model evaluation and adoption
Choose models for different tasks and review usage and request behavior as you develop and introduce new AI features.
About Token Hub
Built for technology teams in Japan.
Started in 2024, Token Hub is a bootstrapped venture led by a solo founder. We operate in Japan and build an enterprise AI gateway primarily for small and medium-sized technology companies in Japan.
Our underlying infrastructure engine, Codex Gateway, supports self-hosting. Our goal is to simplify the connections, access controls, and operations that teams need when working with multiple AI models.