other / agent
StarNet
A desktop harness for creating AI agents and arranging them in a pixel-art space station, with your own model key or a local Ollama model.
At a glance
- Purpose
- A desktop harness for creating AI agents and arranging them in a pixel-art space station, with your own model key or a local Ollama model.
- Runs in
- Local agent with shell, Web service
- Operating systems
- macOS, Windows
- Needs
- Runtime: Node.js 18+ and Git for source runs; Needs internet: Not stated; Needs accounts or API keys beyond your agent: No
- Files and data
- Reads files: Yes; Writes files: Yes; Sends data to other services: Yes
- Tool type
- Agent
- Delivery
- From GitHub
- Version
- Latest on the feat/harness-backend branch
- Access
- Code on GitHub
- Compatibility
- According to the README: Windows 10 and 11 (64-bit) with StarNet_<version>_x64-setup.exe, macOS on Apple Silicon (M1 or later) with StarNet_<version>_aarch64.dmg, and macOS on Intel with StarNet_<version>_x64.dmg. The README says to use the native aarch64 DMG on Apple Silicon. The README describes model access through OpenRouter, sign-in with Anthropic, OpenAI or Google accounts, or local models through Ollama. The README lists agent connections to Telegram, Discord, Slack, Signal and Matrix. The README says you can attach MCP servers. The README describes importing agents from OpenClaw or Hermes. The README says Linux is not a supported public release target.
- External costs
- The README describes model access through an OpenRouter key, supported Anthropic, OpenAI or Google account sign-in, or a local Ollama model. The README's Ollama route installs Ollama and downloads a local model. The README gives no provider prices and describes local ledgers for spend, budgets and run history. Your coding agent or model provider may charge for usage.
- Permissions
- The README says tools operate through capability and consent checks, agents have separate workspaces and permissions, and placed station objects grant capabilities. The README describes using your OpenRouter API key or signing in with supported Anthropic, OpenAI or Google accounts, with keys held in the OS keychain. The README also describes MCP servers, key or OAuth connectors, and connections to Telegram, Discord, Slack, Signal and Matrix. The README says Night Shift continues work within an adjustable limit and logs away actions, and describes cron scheduling. The README says agent memory, transcripts, spend records, tasks and schedules persist on disk, with deliverables written to OUTBOX. The README describes importing persona, instructions, memory and model information from an existing OpenClaw or Hermes home. The README's desktop development route installs dependencies with npm ci. The top-level file list includes scripts/, whose contents the README does not describe.
- Usage terms
- MIT License, as stated in the repository. See its LICENSE file for the exact terms. The README says the MIT License covers the code only. It says Andrew Sims owns the StarNet name, the logo, the station artwork and sprites, and the rest of the brand identity, and that the MIT License does not license them. It says forks and derivatives must use their own name, logo and artwork. It says third-party components remain under their original licences and points to NOTICE.md.
- Interface
- desktop
Code on GitHub
androoAGI/starnet
MITA living pixel-art station where real AI agents do real work. Local-first desktop agent harness - bring your own key, watch your crew actually run.
Topics: agent-harness ai ai-agents byok desktop-app game-ui llm local-first
Public GitHub data fetched · 0 days old
View on GitHub ↗What it does
According to the README, StarNet is a local-first desktop harness for creating agents and arranging them in a pixel-art space station. The README says the station represents live runtime state, with rooms as capability-scoped teams, hallways as handoff lanes and placed objects as capability grants.The README describes concurrent agents with separate workspaces, transcripts, memory and bounded permissions. The README lists Telegram, Discord, Slack, Signal and Matrix connections, Night Shift, Task Briefs, OUTBOX deliverables, recipes, skills, cron schedules, MCP connectors, voice and spend ledgers.
The README describes model access through an OpenRouter key, supported Anthropic, OpenAI or Google sign-in, or a local Ollama model.
Who it is for
People who want to run several AI agents on their own computer with their own model key or a local model, and developers who want to run the harness from source.Example tasks
Create an agent with its own class, persona and loadout.Run several agents at once, each in its own workspace with bounded permissions.
Wire an agent to Telegram, Discord, Slack, Signal or Matrix.
Leave the station running in Night Shift and review the logged away actions.
Import an existing OpenClaw or Hermes agent from its on-disk home.