BYO keys & BYO harness
The production integration contract: the four key types, the three agent execution modes (provider, HTTP endpoint, external), and how your own harness plugs into one score → recommend → verify loop.
Docs
Documentation
Everything you need to wire ReasonRank into your stack, gate quality in CI, and verify the savings in production.
Integrate
Framework adapters
Thin exporter wiring for Vercel AI SDK, OpenAI Agents SDK, LangChain/LangGraph, and Pydantic-AI — plus the zero-adapter OpenTelemetry path that works with any stack.
Evaluate
Agents quickstart
Create an agent, attach how it runs, add test cases, and get your first statistically gated model recommendation in an afternoon.
CI gating
Fail pull requests on quality regressions with the same paired-cluster non-inferiority gate that powers Savings recommendations — locally and in GitHub Actions.
Compare
Keep engineering LLM costs competitive
The evaluate → recommend → gate → verify loop, why the comparisons hold up against eyeballing dashboards, and what a defensible savings number looks like for finance.