OpenAGI is an open-source autonomous agent framework developed by AI Planet (aiplanethub), designed to make human-like AI agents accessible to developers of all backgrounds. Its stated mission is to pave the way for open agents and AGI for all. The framework is available on GitHub at github.com/aiplanethub/openagi and as a pip-installable Python package. The core architecture centers on a multi-agent model where specialized Worker agents each carry a distinct role and instructions, coordinated by an admin agent. A TaskPlanner component enables autonomous decomposition of complex queries into ordered sub-tasks. Setting autonomous=True triggers this planning mode; a human_intervene parameter adds safety checkpoints before execution proceeds. OpenAGI supports multiple LLM backends including OpenAI GPT models and Google Gemini, and integrates with search tools such as DuckDuckGo and Tavily. Agents maintain persistent memory across interactions, allowing them to learn from prior task results. The framework targets Python 3.9 through 3.11. Version 0.2.7 added improved human-intervention controls, new action types, and Anthropic Claude LLM integration. Version 0.3.0, released in early 2026, introduced multimodal agentic workflow tools including Luma AI for text-to-video and ElevenLabs for text-to-speech, as well as VectorDB support for long-term memory retrieval. Note: there is a second, separate project also called OpenAGI hosted at github.com/agiresearch/OpenAGI, which focuses on combining LLMs with domain expert models. The aiplanethub project (this entry) is the one matching the tagline about autonomous agents for all developers. Key features: - Multi-agent architecture with admin coordinator and specialized worker agents - TaskPlanner with autonomous sub-task decomposition (autonomous=True flag) - Human intervention checkpoints via human_intervene parameter - Support for OpenAI GPT, Google Gemini, and Anthropic Claude LLMs - Built-in search tool integrations (DuckDuckGo, Tavily) - Persistent memory and VectorDB support for long-term context - Multimodal actions including text-to-video (Luma AI) and text-to-speech (ElevenLabs) as of v0.3.0 - pip-installable Python package supporting Python 3.9-3.11
Open-source (free). No paid tiers identified for the framework itself.