The models perform badly without guidance but we've found the guidance doesn't need a human in the loop (at least for this task). You can distill what makes a good wiki, what information should go in etc. and the models are good enough to do it reliably. That's what took the most time (researching and studying existing wikis)
When you say support do you mean something where you can collaborate with the model to maintain the wiki?
Understanding what is relevant to be documented vs not is definitely tricky and something we've spent a lot of time on. We found being more prescriptive about what should be documented is helpful, we incorporated guidance from diataxis for it (https://diataxis.fr/)
Also the wiki is primarily for your agents and not for people. Although we're working to improve the prose quality and structure so it is well written for humans too.
Back in the medieval age of 2025, I used frameworks like PydanticAI and LangGraph to build agents. Getting those agent systems to behave reliably was often annoying and tricky.
Claude Code and Codex have since become capable general-purpose agents. They each have an SDK, but those SDKs expose different interfaces and inherit their respective platform’s quirks.
While building CodeAlmanac, I wanted one way to work with both harnesses while retaining features such as skills, custom subagents, workflows, sessions, and goals.
So I built Yoke.
Yoke is a Python SDK for defining an agent once and running it through Claude Code or Codex. Instead of building another agent loop and managing message passing, structured outputs, and context compaction yourself, you reuse the agent harnesses that already exist.
It’s still early. Feedback and contributions are very welcome:
When you say support do you mean something where you can collaborate with the model to maintain the wiki?