Catch AI failures first.
Currai uncovers agent failures, policy violations, and user needs.
One quality layer for every agent interaction
Compare what users asked, what agents claimed, and what tools actually did. All from one platform.
Monitor production conversations, model responses, retrieved context, and tool calls to understand what agents do—and why they fail.
Group recurring user goals and frustrations from real production conversations.
Check agent responses against the behavioral and compliance rules you define.
Turn repeated evidence into concrete prompt, tool, and workflow improvements.
Get started in minutes.
Install the Currai skill
Add the Currai coding-agent skill from your AI application repository. No Currai SDK or collector is required.
npx skills add https://github.com/curraiapp/skills --skill curraiPrompt your coding agent
Ask the skill to find the real chat or agent path, instrument it, and verify the connection with one real interaction.
Use $currai to connect this AI application to Currai using the CURRAI_PUBLIC_KEY, CURRAI_SECRET_KEY, and CURRAI_BASE_URL already configured in the server environment. Instrument the real chat or agent path with authenticated native HTTP capture and verify one real interaction. Never print or copy the secret into source code or chat.Hear where agents lose the customer.
Step through a real support call to see how Currai connects every turn with intent, customer sentiment, agent behavior, and tool evidence.
I need to cancel my internet today and waive the fee. It’s been down all week.
Cancellation, fee dispute, and outage complaint appear in one request.
Every plan includes events, retention, and features that scale with you.
Starter
Free to get started.
Includes:
- ✓Up to 1,000 events / mo
- ✓3-day retention
- ✓Intents and sentiment, discovered automatically
- ✓Quality, policy, and compliance violations
- ✓Alerts for silent failures and rising friction
- ✓Actionable suggestions to improve your agent
- ✓Hosted ingestion, storage & dashboards
Pro
For teams shipping to production.
Includes:
- ✓Up to 10,000 events / mo
- ✓14-day retention
- ✓Everything in Free plan
- ✓Full traces and agent observability
- ✓Slack alerts and notifications
Business
Higher volume, longer history.
Includes:
- ✓Up to 1,000,000 events / mo
- ✓30-day retention
- ✓Everything in Free plan
- ✓Full traces and agent observability
- ✓Slack alerts and notifications
- ✓Priority support
Frequently asked questions
Who is Currai for?
Currai is for product and engineering teams that need to learn from high-volume AI conversations without manually reading every transcript.
What can Currai monitor in production?
Currai monitors user stories, intents, violations, errors, tool calls, latency, and cost across production conversations, with trace evidence attached for investigation.
How long does it take to get my first trace?
About five minutes. Add the Currai coding-agent skill or call the two native HTTP capture endpoints. There's no runtime package or collector to install — the first captured session appears immediately.
Can Currai use my existing instrumentation?
Yes. Currai accepts native HTTP capture, Langfuse-compatible ingestion, and OpenTelemetry traces, so teams can keep existing instrumentation while adding conversational product analytics.
How is pricing calculated?
Each plan includes a monthly event allowance. Free includes 1,000 events, Pro includes 10,000, and Business includes 1,000,000; upgrade when your workspace needs more capacity.
How long is my data retained?
Retention follows your plan, so you keep traces for as long as your plan's window allows. You can export or delete your data on demand at any time.





