Frona v2026.5.0 releases as a self-hosted personal AI agent platform with per-principal sandboxing and policy-driven access control.

Frona is a self-hosted AI agent platform built around a single Rust engine and Cedar-based policy language. Users deploy it on their own infrastructure, point it at an LLM provider of their choice (Anthropic, OpenAI, Google Gemini, DeepSeek, Mistral, Cohere, xAI Grok, Groq, OpenRouter, Together, Perplexity, Hyperbolic, Moonshot, Hugging Face, Mira, Galadriel, or Ollama), and create autonomous agents that browse the web, run code, build apps, make phone calls, and delegate to each other.

Security enforces per-principal sandboxing: every actor (agent, MCP server, deployed app, channel) runs as its own principal with its own policies. CLI tool calls, MCP servers, and apps run in sandboxed Linux processes with policy-driven syscall, filesystem, and network filtering. Credentials are stored in a vault (1Password, Bitwarden, HashiCorp Vault, KeePass, Keeper supported) and agents request secrets at the moment they're needed; credentials never enter agent memory or LLM provider traffic.

Built-in agents at install include Assistant, Researcher, Developer, and Receptionist. Agents support persistent memory with automatic compaction, agent-to-agent delegation, and access to tools including browser automation via Browserless, web search via SearXNG or Tavily or Brave Search, sandboxed code execution (shell, Python, Node.js), app deployment with approval gates, outbound phone calls via Twilio, and scheduling. Channels (Telegram and SMS at launch) let the same agent follow users outside the web UI.

Frona is released under Business Source License 1.1, allowing free use and self-hosting; the only restriction is you may not provide it as a service to third parties. The license converts to Apache 2.0 on 2029-02-28.