OpenSquilla
A microkernel AI agent that routes each turn to the cheapest capable model and runs the same loop across a CLI, web UI, and chat channels.
OpenSquilla sends every turn through a local model router that picks the cheapest model able to handle it, combined with persistent memory, a layered sandbox, built-in web search, and on-device embeddings. Every entry point, whether the web UI, CLI, or a chat channel, runs through the same shared turn loop, so tool dispatch, retries, and decision logging behave the same way everywhere.
A pluggable provider layer connects to more than twenty LLM providers, including OpenRouter, OpenAI, Anthropic, Ollama, DeepSeek, Gemini, and Qwen/DashScope, without requiring changes to application code or configuration. It runs on Windows, macOS, and Linux, with desktop installers, release wheels, or a source build.
It is written in Python and distributed under the Apache-2.0 license.