Leffa

Leffa is an open-source, text-based assistant framework that enables developers to build and run customizable AI agents entirely from the command line. It focuses on reproducibility, extensibility, and transparency, making it suitable for technical users who want fine-grained control over their AI workflows. Leffa supports multiple large language models via configurable backends, allowing users to switch between providers or models without changing their core logic. Its configuration-driven design lets you define prompts, tools, and behaviors in a structured way, enabling repeatable experiments and consistent results.

Key capabilities include interactive chat sessions, scripted conversations, and automated task execution through tool calling. Leffa integrates with external APIs and system commands, so agents can retrieve data, manipulate files, or trigger workflows directly from the terminal. It also supports logging and session histories, which help with debugging prompts, tracking changes, and comparing model outputs over time. Typical use cases include rapid prototyping of AI agents, building CLI-based copilots for development or data analysis, and experimenting with prompt engineering in a controlled environment. Because it is source-available on GitHub, Leffa can be inspected, modified, and extended to fit organization-specific requirements and infrastructure.