Skip to content
onfre

Product

An AI agent that answers, acts and knows when to step aside

Onfre sits on your WhatsApp number, your website chat and your support inbox. Here's what happens behind every reply.

knowledge

It learns from what your team already knows

Connect your help center, policy docs, product catalog and past tickets. Onfre reads them, keeps them in sync, and answers only from what it finds there. When a doc changes, the answers change with it.

  • Help center URLs, Notion, Google Drive, Confluence, PDFs
  • Product and order data from Shopify or WooCommerce
  • Past tickets for tone and real customer phrasing
  • Policy documents always win over old tickets
Returns & exchanges policySynced 4 min ago
help.indigoloom.com · 86 pagesSynced 1 h ago
Shopify catalog · 1,412 productsLive
Ticket export · Jan–Oct 202638 replies excluded (outdated policy)

actions

It does things, not just says things

Most support questions end in an action: check an order, start a return, move a booking, pause a plan. Onfre does these directly in your tools, inside the limits you set, and logs every one.

  • Order lookups, cancellations, address changes
  • Refunds and exchanges, with amount limits
  • Appointment booking and rescheduling
  • Custom actions through your own API (Scale plan)

Action log · today

  • 10:42Exchange created#IL-50917 · M → L
  • 10:44Refund issued₹1,299 · Razorpay
  • 10:51Refund held for approval₹6,400 · over limit
  • 11:03Pickup bookedThu · 10 am–1 pm

guardrails

It follows your rules, every time

You decide what Onfre may do alone, what needs a person's approval, and which topics it should never answer. Rules are written in plain English, and every decision can be traced back to the rule behind it.

  • Approval thresholds for refunds and credits
  • Topics that always go to a human (medical, legal, safety)
  • Business hours, VIP customers and priority queues
  • Personal data redacted from logs and analytics
rule Refunds above ₹5,000 → ask Payments team
rule Mentions of symptoms or medication → hand over, urgent
rule Customer asks for a person → hand over immediately
rule Never promise delivery dates the courier hasn't confirmed

handover

It hands over like a good colleague

When a conversation needs a person, Onfre passes it on with a summary, the customer's history and what has already been tried, and tells the customer who's taking over. Handovers go to your Onfre inbox or straight into Zendesk, Freshdesk or HubSpot.

  • Two-line summary at the top of every handover
  • Customer orders, plan and past contacts attached
  • Routing by topic, language or team
  • Onfre learns from how your agent resolves it

↗ Handover to Payments · Meera

Summary
Refund failed 3× in 5 days. UPI ID verified. Customer frustrated.
Why
Refund over ₹5,000 + customer asked for a person
Order
#PW-22841 · ₹6,400 · delivered 18 Oct
Customer told
“Meera will reply here within 15 minutes”

inbox

Your team works in one inbox, with Onfre in it

Every WhatsApp, chat and email conversation lands in one place, whether Onfre resolved it or your team is handling it. Agents can watch Onfre live, step in at any moment, or ask it to draft a reply for them.

  • All channels in one shared inbox
  • Take over any conversation with one click
  • AI-drafted replies your agents can edit and send
  • Or keep working in Zendesk, Freshdesk or HubSpot
PS

Priya S.

Order status · resolved

AI
AK

Arjun K.

Refund ₹6,400 · waiting for you

You
KM

Kabir M.

Pause plan · resolved

AI
NR

Neha R.

Sizing question · Onfre typing…

AI

insights

It tells you what to fix

Every week Onfre shows what customers asked, what it resolved, where it handed over and why, and which questions had no good answer in your docs. Most teams find their biggest product and delivery problems in this report.

  • Top topics and how they trend week to week
  • Missing or contradictory documentation
  • Resolution rate, handover reasons, CSAT
  • Export to Google Sheets or your BI tool

71%

Resolved by AI

9 s

Median first reply

18%

Handed over

6

Doc gaps found

Example report for a fictional store.

Under the hood

Inside a single resolution

Here's what happens in the few seconds between a customer's message and Onfre's reply.

bhai mera order 6 din se nahi aaya 😟

Aaj shaam 7 baje tak pahunch jayega! 📦

  1. Step 1

    Understand

    Works out what the customer wants, in which language, and how they're feeling.

    • Intent: order status
    • Language: Hinglish
    • Mood: worried

  2. Step 2

    Look it up

    Pulls only what this question needs from your docs and systems.

    • Order #IL-48213
    • Courier: out for delivery
    • Shipping policy §1

  3. Step 3

    Check the rules

    Applies your guardrails: confidence, blocked topics, approval limits.

    • Confidence 0.95
    • No approval needed
    • No blocked topic

  4. Step 4

    Act

    Does whatever the answer needs, inside the limits you set.

    • Subscribed to delivery alert

  5. Step 5

    Reply

    Answers in the customer's language and your brand's tone.

    • Hinglish · friendly
    • Tracking link added

  6. Step 6

    Learn

    Logs the sources and outcome, and feeds your weekly report.

    • Marked resolved
    • Topic counted

One agent, every channel

WhatsApp

Official Business Platform. Templates, media, buttons and order messages.

Web chat

A light widget for your site, styled to your brand, live in one line of code.

Email

Forward your support inbox. Onfre replies in-thread and keeps your signature.

Instagram DMs

Answers product and order questions straight from your DMs.

Messenger

Same answers and actions as every other channel.

SMS

For order updates and customers who don't use chat apps.

anyone there? it's 2 a.m. 😅Always. How can I help?

See it answer your own tickets.

Connect WhatsApp, point Onfre at your help docs, and see it draft answers to your real tickets before any customer does.

Start free trialBook a demo