Agent Sandbox
Most teams deploy their AI agent and find out it's broken from a customer complaint. Sandbox gives you a live environment to simulate conversations, inspect every decision, and ship with confidence.
Know exactly how Rian will respond. Before it matters.
Deploying an AI agent without testing it is a risk no support team should take. Wrong answers, missed escalations, and incomplete context all damage trust and you only find out after a real customer experienced it.
Sandbox puts you in control before go-live.
Three views. Full visibility.
Session
Run real conversation simulations against your live agent configuration. Type as a customer, choose a test persona new visitor, returning user, frustrated customer, enterprise account and watch Rian respond exactly as it would in production.
The same knowledge base. The same escalation rules. The same AI.
Use Session to verify Rian answers your most common support questions correctly, handles edge cases without looping, and escalates when it should.
Events
Every response Rian generates goes through a series of decisions classifying intent, retrieving knowledge, selecting a response flow, generating a reply. Events surfaces the full log of that process in real time.
See which intent class was detected, what confidence score the classifier returned, which knowledge base article was matched, which model tier handled the request, and how long each step took.
When something responds unexpectedly, Events shows you exactly why.
Memory
Rian draws on multiple layers of context to generate a response. Memory lets you inspect each one.
Hot context: Active session state and confirmed intent.
Warm cache: Account data and recent conversation history.
Cold storage: Knowledge base articles and scraped website content, shown with the relevance score that triggered each match.
How it works
- 1.
Rian scrapes your website on setup. Your product content, support articles, and configuration become the foundation of what Rian knows.
- 2.
Open Sandbox. No additional setup required. Sandbox uses your live agent configuration automatically.
- 3.
Simulate a conversation. Type as a customer or choose a test persona and watch Rian respond in real time.
- 4.
Inspect the result. Switch between Session, Events, and Memory to see exactly how Rian processed the request.
- 5.
Go live when you're ready. One click. Your embed script is ready the moment you activate.
Built for founders who ship fast and care about quality.
Renprofile is built for SaaS teams that don't have a dedicated support ops function but still need customer support to work properly from day one.
Sandbox exists because deploying AI blind is not an option when your customers are the ones who discover the mistakes.
Available on all plans. No additional setup required.
Frequently asked questions
Quick answers to common questions about the Agent Sandbox.
Yes. Sandbox runs against your actual agent config — the same knowledge base, escalation rules, and model settings Rian uses in production. What you see in Sandbox is what your customers will experience.
No. Sandbox sessions are flagged separately and excluded from all production metrics and resolution rate calculations.
Yes. Sandbox includes test personas so you can simulate different conversation styles without writing every message manually.
When Rian answers your most common support questions correctly, handles a frustrated customer without looping, and escalates when you'd expect it to. Sandbox gives you the visibility to make that call with confidence.
Yes. Sandbox is available to every Renprofile user at no additional cost.