
A self-hosted AI traffic control

A self-hosted AI traffic control
GreyFox is a self-hosted tool designed to intercept and route local AI application traffic, enabling operators to inspect, debug, and enforce policies on large language model data streams. Developed by Skilful Fox Studio, it functions as a local control plane that keeps operational data within the container boundary, avoiding any telemetry or hosted surveillance layer. The tool is intended for users who need to manage AI traffic flows without sending data to external cloud services, supporting both individual developers and small teams. The tool offers a range of features centered on local proxy inspection, manual route overrides, and single-profile storage. Its Community Edition provides a baseline set of capabilities, while paid Essential and Pro editions add commercial workflow features, including up to 25 active managed users, up to five saved environment profiles, and up to ten deterministic request/response presets. Advanced analytics for tokens, cache, traffic sources, models, guards, and latency are available in the Essential tier, along with enterprise audit log exports and custom routing policy scripts using JavaScript or WebAssembly. Configurable limit and quota schedules are also supported. The interface is designed as a self-hosted control surface, offering a local overview of traffic history, a managed-user control surface, and a view of known managed users with their operational state. Manual informational estimates are kept separate from authoritative token monitoring. Typical use cases include managing AI traffic for local operator workflows, controlling which provider routes are used, changing the base URL while preserving the rest of the client flow, and enforcing policies on LLM interactions. The tool is suited for environments where data privacy and local control are priorities, such as research labs, development teams, or organizations that require telemetry-free operation. The software is distributed as a containerized application, and paid subscriptions are processed through Stripe with repository access managed via GitHub invites. The Community Edition is currently at version 0.1.1, with a maintenance patch, while Essential 0.2.1 is available for paid access. The Pro edition is in development, with advanced commercial boundaries being shaped after the Essential release.