StarNet – Visual Desktop Hub for Real AI Agents

What it is

StarNet is a local‑first desktop application that lets you create, organise and run multiple AI agents in a pixel‑art “space‑station” UI. Each room in the station represents a team with its own permissions, memory and budget, and the visual layout is the actual workflow the agents execute. The app talks to real LLM providers (OpenAI, Anthropic, Google, OpenRouter, or a local Ollama model) and can call external tools, so the agents do real work – generating files, sending messages, running scheduled jobs, etc.

Core features

Feature	What you get
Multiple distinct agents	Separate personas, load‑outs, workspaces, memory, transcript logs and permission sandboxes.
Bring‑your‑own models	Use cloud keys (OpenRouter, OAuth to Anthropic/OpenAI/Google) or a local Ollama model; keys are stored in the OS keychain, never in the UI.
Channel integration	Connect agents to Telegram, Discord, Slack, Signal or Matrix so you can interact with the station from anywhere.
Night‑shift / away mode	Agents keep running under a configurable leash; every action is logged for later review.
Recipes, skills & schedules	Pre‑built multi‑step workflows, reusable skill modules and cron‑style scheduling with visible output.
Ask‑before‑guess	Ambiguous tasks are turned into a concrete question with selectable options, avoiding silent wrong guesses.
Outbox for deliverables	Finished work is saved as real files in an OUTBOX folder rather than hidden in chat history.
Model Context Protocol (MCP) connectors	Plug‑in servers that expose additional capabilities to agents via a standard protocol.
Voice I/O	Push‑to‑talk dictation and a consistent station voice for responses.
Transparent ledgers	Persistent on‑disk logs of spend, budgets and run history; the UI never shows state it cannot prove.

How it works

• Frontend – vanilla JavaScript renders the pixel‑art station and UI. It receives live events from the backend via HTTP/NDJSON and Server‑Sent Events.
• Sidecar – a Node.js runtime that handles provider authentication, model calls, tool execution, persistence, consent checks and budgeting. Secrets stay in the OS keychain.
• Desktop shell – built with Rust/Tauri, bundling the frontend and sidecar into native Windows/macOS installers.
• Persistence – agents’ memory, transcripts, budgets and schedules are stored on disk, enabling offline‑first operation.

Getting started

1. Download the installer for Windows or macOS from the StarNet releases page.
2. Run the app, add a provider (OpenRouter key or OAuth sign‑in) or install Ollama locally for a free model.
3. Create your first agent, place rooms and objects in the station, and watch the agents act in real time.
4. Optional – run from source: clone the repo, start the Node sidecar (node sidecar/index.js), then open http://localhost:8787.

Who it’s for

• Developers or power users who want a visual, state‑driven interface for orchestrating multiple LLM agents.
• Teams looking for a local‑first, privacy‑preserving way to run AI workflows without cloud‑only platforms.
• Hobbyists who enjoy a pixel‑art UI while their agents actually read/write files, call APIs, and interact on chat platforms.

License & branding

The code is released under the MIT License. The StarNet name, logo and artwork are trademarked by the author, so forks must re‑brand.

---

All information above is taken directly from the repository’s README.