ThinkWatch
An enterprise AI gateway that proxies OpenAI, Anthropic, Gemini and self-hosted LLM APIs plus MCP tool calls behind one control plane, with RBAC, audit logs, rate limiting and cost tracking.
ThinkWatch is a gateway meant to be the single path through which all AI API calls and MCP tool invocations in an organization flow, similar to how an SSH bastion host is the single path for server access. It natively serves OpenAI Chat Completions, Anthropic Messages, OpenAI Responses and Gemini-format APIs on one port, converting between formats automatically, and routes to OpenAI, Anthropic, Google Gemini, Azure OpenAI, AWS Bedrock or any OpenAI-compatible endpoint.
It issues scoped virtual API keys that work across both the AI gateway and an MCP gateway through a per-key allowlist, streams responses with real-time token counting, and manages key rotation, inactivity timeouts and expiry warnings. It is aimed at giving an organization visibility into which model and MCP tool each request used, at what cost, replacing scattered hardcoded API keys with a single audited, access-controlled deployment.