CoreLayer Documentation
Start here to install CoreLayer, configure a model provider, connect your first MCP server, and understand the permission model before giving Jarvis access to tools.
Welcome to the CoreLayer documentation. CoreLayer is a local-first desktop AI control layer that connects your personal apps, tools, models, and MCP workflows through a permission-guarded command center.
Quick Links
Install CoreLayer Get the desktop app running on your system.
Quickstart Connect your first model and MCP server in 5 minutes.
Permission Guard Understand how tool execution stays safe and auditable.
Architecture Learn how the Tauri shell, daemon, and packages work together.
What is CoreLayer?
CoreLayer sits between you and your personal app ecosystem. Instead of a chatbot trapped in a browser tab, it operates as a desktop control layer:
- Voice-native — wake word, streaming ASR/TTS, barge-in interruption
- MCP-first — connect external tool servers through the Model Context Protocol
- Tool-aware — unified registry for native modules, MCP tools, skills, and REST adapters
- Permission-guarded — risky actions are classified, paused for approval, and logged
- Local-first — SQLite on your machine, optional Supabase cloud sync
Documentation Sections
| Section | What you'll find |
|---|---|
| Getting Started | Install, quickstart, first MCP connection |
| Configuration | Model providers, MCP servers, storage, voice, permissions |
| Core Concepts | Jarvis, Control Center, Model Gateway, Tool Registry, Permission Guard |
| Guides | Step-by-step guides for common workflows |
| Reference | Config file, CLI commands, API endpoints |
| Architecture | System design, runtime boundaries, data flow |