Agentlas OS
An Apache 2.0 AI agent framework that indexes documents and searches memory locally, coordinating specialists through your chosen model and host.
Agentlas OS is an open-source AI agent framework for projects that need specialist teams and shared project context. It runs on macOS, Windows and Linux, with local document indexing, semantic search and memory retrieval. You supply the model. Using a cloud model sends requests to that provider and may incur charges.
Its document engine parses local files into a SQLite database and uses a bundled Model2Vec model for semantic retrieval without a hosted embedding API. It also parses Korean HWPX and HWP5 documents. Project memory and each agent's experience have separate storage boundaries, with privacy checks and secret redaction before context reaches the host.
The framework divides a goal into tasks with defined roles, deliverables and handoffs. It finds candidates among installed agents, private Agent Cloud packages and public Agentlas Hub agents. Your model selects specialists and records its reasons; selected releases stay pinned to their exact contents. A separate verifier checks the combined result before the framework marks work complete.
The Apache 2.0 core works inside Claude Code, Codex, Cursor and Gemini, where the host controls tools and permissions. Agent Cloud stores and retrieves packages; it doesn't perform model inference, and packages don't transfer credentials or local files between computers.
The separate Agentlas Desktop app runs the same runtime and adds local LLM access through llama.cpp with GGUF models, LM Studio and MLX.
