This project is scheduled for launch
Launch date: Tuesday, August 18, 2026 at 08:00 AM UTC
Scheduled
Brief Introduction
GreyFox, from Skilful Fox Independent research studio, is a self-hosted, telemetry-free AI traffic tooling solution designed for local operator workflows. It enables users to inspect, debug, and enforce policies on LLM data streams within their own Docker environment, ensuring complete control and data privacy.
It is ideal for developers, AI engineers, and organizations that require granular control and privacy over their local AI application traffic and LLM interactions.
Key Features
- Self-hosted, telemetry-free control plane for AI traffic.
- Intercepts and routes local OpenAI-compatible LLM application traffic.
- Inspect, debug, and enforce policies on LLM data streams.
- Local operator panel with traffic history, user control, and provider settings.
- OpenAI-compatible proxy integration for seamless client-side adoption.
- Community Edition for baseline inspection and simple proxy behavior.
Use Cases
Secure AI Development & Testing: Developers can use GreyFox to safely experiment with AI models and LLM integrations locally, inspecting traffic and debugging interactions without exposing sensitive data to external cloud services or telemetry. This ensures privacy and compliance during the development lifecycle.
Controlled AI Application Deployment: For organizations deploying AI applications, GreyFox provides a robust control surface to manage routing, enforce usage policies, and monitor traffic within a self-hosted environment. This is particularly valuable for maintaining data sovereignty and operational oversight for multiple users and providers.
Bridging Research to Production: GreyFox helps transition AI research utilities into productized operator workflows by offering a consistent and controlled environment for managing AI traffic. It allows for repeatable setups and advanced analytics, moving beyond scattered configurations to a centralized management system.
Pricing Information
GreyFox offers a freemium model, including a public, open-source Community Edition for basic local AI traffic inspection. Paid Essential subscriptions are available monthly (EUR 19) or annually (EUR 179) via Stripe, providing advanced features like managed users, environment profiles, and deep traffic analytics. A Pro edition with enterprise-grade features is currently in development.
User Experience and Support
GreyFox features a local operator panel designed for intuitive management of AI traffic workflows, emphasizing a self-hosted control surface over a surveillance layer. The Community Edition 0.1.1 includes UI template and styling refinements for ease of use. Support is available via support@skilful-fox.com, and discussions for paid editions are encouraged.
Technical Details
GreyFox operates as a local proxy server, integrating with OpenAI-compatible providers by simply changing the base URL in client applications. It runs within a Docker environment, ensuring self-hosting and isolation. The package is distributed via GitHub Container Registry, with TypeScript examples provided for integration.
Pros and Cons
- Pros:
- Ensures data privacy and sovereignty with a telemetry-free, self-hosted architecture.
- Provides granular control and visibility over local AI application traffic.
- Easy integration with existing OpenAI-compatible LLM clients.
- Offers a free Community Edition for foundational use.
- Supports advanced commercial workflows with paid editions.
- Cons:
- Requires familiarity with Docker for self-hosting.
- Advanced policy and audit features are still under development in the Pro edition.
- Cost calculations are manual estimates, not authoritative token monitoring.
Conclusion
GreyFox stands out as an essential tool for anyone building or deploying AI applications who prioritizes privacy, control, and local operational oversight. Its unique self-hosted, telemetry-free approach empowers users to manage LLM data streams effectively and securely. Explore GreyFox today to gain unparalleled control over your AI workflows.
Comments
Comments will be available once the project is launched.
