Case studies
Hosted AI platforms keep shipping new features: a design canvas, a research agent, a coding assistant that opens pull requests. Each case study takes one of those features and shows how we built the same thing on OAP from primitives the platform already has, running in your own cluster, with every action checked before it runs.
How each case study is laid out
Every case study answers the same four questions, in the same order:
- What is it? What the feature does, and what makes it useful.
- The ingredients. What any team would need to build its own: a place to talk to it, a place for its output, access to the right data, and the guardrails around all three.
- How we built it in OAP. The actual agent definition, start to finish, with the parts that matter shown in full.
- What's different on OAP. Where the OAP version is more secure, more portable, or easier to adapt — and where the hosted original is still ahead.
The series
- We built our own Claude Design: how we built messagingbot, a brand designer that turns a brief into an on-brand page and revises it from notes pinned to the design, without writing a line of new platform code.