Overview
AutoGPT is an autonomous AI-agent framework and platform that chains large-language-model reasoning, tool invocation (web browsing, code execution, file I/O) and vector-store memory to achieve goals expressed in natural language. What began as a proof-of-concept for GPT-4 has grown into a full cloud- and self-hosted platform that can run persistent agents for business and personal automation.
Key Capabilities
- Goal-oriented task decomposition that breaks objectives into sub-tasks and executes them unsupervised.
- Extensible tool system for web search, APIs, code, file access and custom plugins.
- Long-term memory via vector stores so agents learn from previous runs.
- Multi-model support (OpenAI GPT-4/3.5, local Llama models via Ollama, Groq, etc.).
- Workflow builder & marketplace for sharing reusable “blocks” and entire agents.
- Dual-license approach: MIT for the classic agent; Polyform Shield for new platform components.
