The blog post includes link to the repos (MIT Licensed), a video of the agent in action, sample code produced, and an Emacs org-mode literate programming document explaining the implementation and tests. There are links to an interactive playground for the sw-MLPL language.
The git repo can be used to run this agent locally on a system with Ollama and a coding LLM. Does not yet connect to cloud LLMs.
I hope this helps someone learn how a coding agent can be built. Inspired by (but not ported from) OpenCode.