Docs
SDKs for Python, TypeScript, Go and Rust, a REST API with a live event stream, webhooks, a GitHub App and an MCP connector for agents. Below is the exact call behind a real run, and what came back.
Quickstart
Every SDK has the same shape: create a run, stream it, read the verdict. The code below is straight from each SDK's readme.
The stream
A run streams as it happens, over server-sent events: each persona as it's seated, each reaction as it lands, and the report the moment the last one does.
persona.readyA seat is built: archetype, voice and decision profile.
reaction.partialA reaction arriving word by word, for live UIs.
reaction.completeOne persona's final read, with its sentiment and objections.
sim.completeThe report: the verdict, the split, the objections, the quotes.
Where it plugs in
On a pull request, at your own endpoint, inside an agent, or in the channel your team already reads.
MCP
Add one connector and the assistant you already use can put copy, ads, pages and decisions in front of a room, wait for the verdict, and tell you what it found in plain words.
https://mcp.kranth.ai/mcp as a connector in your agent.The reference
The full reference is interactive: every endpoint with its fields, errors and examples, runnable against your own key.
Your key
Sign up, mint a kr_live_ key under API keys, and every account starts with 10 free credits, no card.