opensquilla

OpenSquilla is a token-efficient AI agent built on a microkernel architecture, designed to maximize capability within a fixed budget. It features a local model router that dynamically selects the most cost-effective model for each interaction, while supporting persistent memory, a layered sandbox, built-in web search, and on-device embeddings. All entry points (Web UI, CLI, and chat channels like Telegram, Slack, or Discord) share the same turn loop, ensuring consistent behavior for tool dispatch, retries, and decision logging. The system integrates with 20+ LLM providers (e.g., TokenRhythm, OpenRouter, OpenAI, Anthropic, Ollama, DeepSeek, Gemini, Qwen) via a pluggable provider layer, requiring no code or config changes. OpenSquilla prioritizes privacy with configurable usage statistics and supports migration from OpenClaw or Hermes Agent. Installation options include desktop apps, terminal-based setups, or source-based development, with cross-platform support for Windows, macOS, and Linux.