OpenSquilla

OpenSquilla is a token-efficient, microkernel AI agent that runs on your CLI, Web UI, and chat channels. A local model router sends each request to the cheapest model that can handle it, while persistent memory, a layered sandbox, built-in web search, and on-device embeddings work together in a single shared turn loop. Every entry point uses the same loop, so tool dispatch, retries, and decision logging behave identically everywhere. A pluggable provider layer connects to over 20 LLM providers including OpenAI, Anthropic, Ollama, DeepSeek, Gemini, and Qwen with no code changes. Key features include the intelligent SquillaRouter for multi-model routing, persistent memory with sandboxed execution, on-device embeddings, built-in web search via Brave and OpenRouter, MCP protocol support, and a beautiful Vue-based Web UI control console. OpenSquilla runs everywhere — as a desktop app on Windows and macOS, as a terminal CLI, or via Docker on Linux — with the same feature set across all platforms.