Fovux MCP

From dataset to deployed ONNX, in one conversation.

Fovux MCP is the local Python server behind the Fovux workflow. It exposes MCP tools covering:

• dataset inspection, validation, duplicate detection, conversion, and splitting
• non-blocking YOLO training lifecycle management
• evaluation, per-class reporting, and error analysis
• export, quantization, inference, and latency benchmarking
• run and model comparison for day-to-day iteration

What Makes Fovux Different

• Local-first by default: no SaaS account, no cloud lock-in, no hidden control plane
• Built for edge-CV practitioners, not generic automation demos
• File-oriented state under ~/.fovux/ so runs, exports, and reports stay portable
• Designed to pair with MCP clients and with the Fovux Studio VS Code extension

Start Here

1. Read Getting Started for the 5-minute tour.
2. Learn the core mental model in Concepts.
3. Review Testing and Code Coverage for our CI and coverage policy.
4. Review MCP 2025-11-25 Conformance for protocol and transport support.
5. Jump to Tool Reference for exact inputs, outputs, and examples.

Availability

Fovux is currently installed from source from this repository. The documentation in this directory is the canonical reference for the current release.