nano-agent lets you run GGUF‑quantized LLMs on low‑power hardware using the llama.cpp backend, exposing a simple CLI and REPL. It includes a plugin system for tools (bash, Python, WASM) that the agent can invoke via subprocesses. Designed for developers who need autonomous agents on edge devices with limited RAM and no internet connectivity, it offers a fully offline alternative to cloud‑centric frameworks. Its lightweight footprint and hardware‑specific optimizations make it more practical for embedded AI than generic agent libraries.
View on GitHub →Nemo-Forge/nano-agent