ThinkWatch

Why we included this project

Once a company starts pointing multiple agents, coding assistants, and internal tools at the same set of model providers, the API keys and usage numbers get messy fast. ThinkWatch gives you a single gateway in front of OpenAI, Anthropic, Gemini, Azure OpenAI, AWS Bedrock, and self-hosted models, and it also handles MCP tool calls, so you can authenticate users, enforce role-based access, rate-limit, and keep an audit log in one place. The management console shows who called which model, how many tokens they used, and what that cost, which is the kind of accounting that matters when keys are scattered across teams. If you're the person responsible for governing AI usage rather than just calling models, this is a control plane worth a look.