Documentation
Welcome to the Wylde docs. Wylde is a fully native AI platform for Windows — twenty cooperating services, one desktop app, all running on your machine.
Install the Fletch app, launch the platform, and run your first health check in under five minutes.
How the named-pipe IPC, discovery layer, gateway, and twenty services fit together.
A reference for every service that ships with Wylde — ports, named pipes, and what each one does.
All environment variables — voice, RAG, trainer, vault, VPN, captioner — with defaults and examples.
What Wylde is
Wylde is a desktop AI platform that bundles inference, retrieval, training, voice, automation, and graph storage into a coordinated set of native services. The Fletch GUI (a Tauri 2 + Svelte 5 desktop app) is the front door, but every service also exposes a Windows named pipe so other tools — including command-line scripts and your own integrations — can talk to it directly.
The whole platform is designed around three properties:
- Local-first. All twenty services bind to
127.0.0.1. Nothing is sent off-device unless you explicitly turn it on. - Composable. Services self-register, tools are short Python files, workflows are YAML you can read.
- Fast. Hot-path IPC over named pipes runs at ~0.1 ms — about 15× faster than HTTP loopback.
Where to next?
If this is your first time here, head to Getting started. If you want to know how the system is laid out, the Architecture page has the full diagram.