L2MAC

L2MAC is a framework for a general-purpose stored-program "automatic computer" implemented as an LLM-based multi-agent system. From a one-line input prompt, it generates large outputs such as an entire complex codebase. Internally, it creates and executes a prompt program made of a series of instruction steps on its own; each step is handled by a new LLM agent, and a control unit manages the context. Agents read and write persistent memory in the form of a file store, and can use tools that run syntax checks and unit tests. The related paper was accepted at ICLR 2024.

• Work around context limits through collaboration of multiple LLM agents
• Generate large outputs such as codebases and books from a single prompt
• Memory that stores and uses past interactions and outputs
• Automatic generation and execution of sequential prompt programs
• Integration with syntax-check and test-execution tools
• Customize task execution steps

Open source (free)

Mostly distributed free under open-source licenses; the cost of the LLM API you connect to is separate

Estimate based on AI knowledge - please verify on the official site

• One of 36 AI tools in the Agent Frameworks field.
• Connected to 5 other tools — #11 in its field.