I ran OrchestKit — my open-source Claude Code toolkit — through ORA, an independent audit that scores how agent-ready a product is: how easily an AI agent can discover you, identify itself, authenticate, and actually get work done against your surface.
Result: 83/100, Grade B ("Competitive"), ranked #49 of 12,200.
The breakdown
| Layer | Score | |
|---|---|---|
| Auth & Access | 28/30 | 🟢 Strong |
| Agent Integration | 18/20 | 🟢 Strong |
| Identity | 18/20 | 🟢 Strong |
| User Experience | 10/10 | 🟢 Perfect |
| Discovery | 9/20 | 🟠 Partial |
What I take from it
The foundation is strong: an agent can identify, connect, and use OrchestKit with near-zero friction — exactly what an agent-native toolkit should be.
The one hole is Discovery — presence, not product. The thing is ready; it just needs more people (and agents) to know it exists. Honestly the most fun problem to have, and it's the next focus.
Solo-built. Open-source. Independently verified.
Top comments (1)
Agent readiness should measure whether a tool can be used safely by another system, not whether the demo path looks smooth. The details that matter are schemas, errors, idempotency, auth, rate limits, and documentation an agent can actually reason over.
That is why readiness scores are interesting if they expose gaps. A score without the failure categories would be much less useful.