Getting started

Introduction

Understand what Currai learns from real AI conversations and how to turn that evidence into product improvements.

Currai is product intelligence for AI applications. It watches real conversations and agent runs, explains what users are trying to accomplish, detects behavior that breaks your policies, and keeps the source evidence attached so your team can act with confidence.

What Currai helps you answer

  • What do users need? User Stories reconstruct the job behind each conversation. Intents group recurring goals such as pricing questions, cancellation requests, or failed setup attempts.
  • Where is the agent failing? Violations compare agent output with the policies you define. Errors surface failed model, tool, and workflow operations.
  • What needs attention now? Alerts watch durable conditions. Pro and Business workspaces can also receive Daily Slack Reports with the biggest conversation patterns, errors, and policy violations.
  • What should we improve next? Command Center summarizes the workspace, while Auto Improve turns repeated production evidence into proposals your team can review.

The workflow

  1. Connect the real chat, agent, or MCP path with the Currai Skill.
  2. Run one real interaction and confirm it appears in Events and User Stories.
  3. Define the intents and violation rules that matter to your product.
  4. Inspect matched evidence and linked conversations.
  5. Configure dashboard alerts or a Daily Slack Report.
  6. Use Command Center and Auto Improve to prioritize product changes.

Currai also preserves the nested model, tool, retrieval, guardrail, and evaluator steps behind a conversation. Those traces are supporting evidence: open them when you need to understand why a signal appeared or where a run failed.

Start here