Prism-Shadow/penguin-harness
PenguinHarness pitches a local-first system where agents generate, evaluate, optimize, and deploy other agent apps.
The project says a one-sentence prompt can produce a runnable agent application, citing a RAG app example it says cost $0.02 in DeepSeek V4 Pro tokens. It claims its minimal toolset lowers token use and that its data-analysis accuracy beat Claude Code at 1/70 of the cost in its paired-model benchmark. The repo also lists desktop and CLI installs, built-in plugin categories, support for 1000+ local and online models through presets or OpenAI-compatible endpoints, and an Apache 2.0 license. Source: GitHub · LLM repos' note.
The project says a one-sentence prompt can produce a runnable agent application, citing a RAG app example it says cost $0.02 in DeepSeek V4 Pro tokens. It claims its minimal toolset lowers token use and that its data-analysis accuracy beat Claude Code at 1/70 of the cost in its paired-model benchmark. The repo also lists desktop and CLI installs, built-in plugin categories, support for 1000+ local and online models through presets or OpenAI-compatible endpoints, and an Apache 2.0 license. Source: GitHub · LLM repos' note.
score 4