Documentation is under construction

This page is under construction and will be available in the future.

GET STARTED

Run a local-first AI workspace your organization can operate.

Obrenna runs its control plane, agent runtime, registry, policies and local state on the desktop. Remote model or MCP services are optional and explicitly configured.

Before you begin

Use the current desktop developer setup, a local Ollama installation or another OpenAI-compatible endpoint, and any MCP servers you intend to connect. Public installer, enrollment-code and remote control-plane workflows: NA.

Current architecture

1Tauri desktop shell

Owns local process lifecycle, permissions and the desktop UI.

Implemented

2FastAPI backend

Handles orchestration, MCP registry and routing, policies, confirmations and local APIs.

Implemented

3Local state and runtime

SQLite, local files, memory and JSONL telemetry remain on the machine by default.

Implemented

4Optional remote services

Prompts, responses, tool arguments and results leave the machine only for explicitly configured remote endpoints.

Available

Current local setup examples

# Current developer/setup flow
dev.ps1

# Install the local browser dependency when needed
cd backend
python -m playwright install chromium

# Pull example local models
ollama pull qwen3.5-9b-claude-opus-reasoning-distilled
ollama pull granite4.0-h-micro-3b
ollama pull qwen3.5-0.8b

NextUnderstand the local control plane and data boundary