osaurus
View on GitHubOwn your AI. The native macOS harness for AI agents -- any model, persistent memory, autonomous execution, cryptographic identity. Built in Swift. Fully offline. Open source.
Osaurus is a native Swift/SwiftUI macOS harness for AI agents: local or cloud models, persistent memory, sandboxed code execution, browser and computer use, image generation, MCP support, and cryptographic agent identity. Fully offline-capable and MIT licensed.
Use Cases
Run autonomous AI agents entirely offline on a MacGive agents persistent 3-layer memory and file/search/git toolsExecute agent code in an isolated Linux VM sandboxLet agents browse and operate websites with per-agent sessionsGenerate images locally from text promptsScrub PII and secrets before sending prompts to cloud modelsExpose local agents to the internet over encrypted relay channelsSpawn subagents on local or cloud models with budgetsServe as an MCP server/client for other AI toolingDelegate tasks between agents with end-to-end encrypted channels
Built With
- Language
- Swift
- Frameworks
- MLX · Apple Foundation Models · Apple Intelligence · Apple Containerization · WebKit · SwiftUI · SQLCipher · MCP · OpenAI API · Anthropic API · Ollama API
Tags
macos · swift · apple-silicon · local-llm · offline · multi-agent · mcp-server · persistent-memory · sandbox · computer-use · browser-automation · privacy · image-generation · orchestrator · subagents · cryptographic-identity