New:Ravan Revenue Recovery — voice AI priced on recovery, not on calls →
Product · Knowledge base

Train agents on your company knowledge.

Let agents answer from approved company content instead of generic AI guesses. Build knowledge bases from text, URLs, and files — agents search them live during every call.

Book a demoView API docs
Text, URL & file sourcesAuto-refreshed URLsRetrieval playgroundLive retrieval on calls
app.ravan.ai/knowledge-bases/support-kb

Knowledge base

Customer support — policies & pricing

5 sourcesConnected to 3 agents
pricing-plans-2026.pdfFile—Indexed
ravan.ai/pricingURLAuto-refresh onIndexed
Refund & cancellation policyText—Indexed
help-center/faqURLAuto-refresh onIndexed
product-catalog-q2.csvFile—Indexing…

Illustrative product view

How it works

From your content to grounded answers.

A knowledge base is the approved source of truth behind each agent. Add sources, index them, test retrieval, and connect the result to live calls.

Text sources

Paste policies, scripts, FAQ answers, and process notes directly. Fastest way to give an agent the exact wording your team has approved.

URL sources

Point a knowledge base at your website pages — pricing, help center, product pages — and Agni indexes the content for retrieval.

File uploads

Upload documents like catalogs, rate cards, and policy PDFs. Indexed content becomes searchable during live conversations.

Auto-refresh

Enable auto-refresh on URL sources so the knowledge base stays current when the underlying page changes — no manual re-uploads.

Retrieval playground

Ask test questions in the playground and inspect what the agent would retrieve — verify answers before any customer hears them.

Live retrieval on calls

Agents search connected knowledge bases in real time during calls, grounding answers in your content instead of generic AI guesses.
What to build

One knowledge base per kind of truth.

Keep content organized the way your business already thinks about it. Separate knowledge bases keep retrieval precise and updates simple — and each agent can be assigned the ones it needs.

  • Product knowledge — specs, features, availability, and catalog details
  • Policy content — refunds, warranties, eligibility, and terms
  • Pricing — plans, rate cards, and offer conditions
  • FAQ and support content — the answers your team already trusts
  • Internal process notes — escalation rules and operating procedures

Knowledge base limits vary by plan — 2 on Starter, 5 on Growth, 15 on Scale. Compare plans.

Deployment path

Test retrieval before customers hear it.

Knowledge bases follow the same discipline as agents: nothing reaches a live call until your team has verified it.

01

Create a knowledge base

Set up separate knowledge bases for product, policy, pricing, FAQ, or internal process content.

02

Add and index sources

Add text, URLs, and files. Agni indexes the content and keeps auto-refreshed URL sources current.

03

Test in the playground

Run retrieval queries and check the answers before connecting the knowledge base to an agent.

04

Connect to agents

Assign knowledge bases to agents. On live calls, agents search them in real time to answer accurately.

Knowledge is one half of resolution — tools are the other. Agents can also transfer calls, trigger functions, and hand off to humans. See the agent builder or explore APIs & webhooks.

Where it matters

Grounded answers change support economics.

Knowledge-backed agents handle the questions that make up most inbound volume — policies, pricing, order status, eligibility — and escalate the rest with context.

Customer support

Answer FAQs, explain policies, and collect context before escalating complex calls to human teams. Support use case →

Sales & qualification

Keep product, pricing, and offer details accurate while agents qualify leads and handle objections. Sales use case →

E-commerce

Ground COD confirmations and delivery follow-ups in current catalog, policy, and order information. E-commerce use case →
FAQ

Knowledge base questions, answered.

Can agents access a knowledge base?
Yes. Each agent can be assigned multiple knowledge bases — uploaded documents, FAQs, product specs, pricing tables, and more. Starter plan supports 2 knowledge bases; Growth 5; Scale 15. The agent retrieves relevant context in real time during the call.
What kinds of sources can a knowledge base hold?
Three source types: pasted text (policies, scripts, approved answers), URLs (your website, help center, pricing pages), and uploaded files (catalogs, rate cards, policy documents). All sources are indexed so agents can search them during live conversations.
Do URL sources stay up to date?
Yes, when auto-refresh is enabled. Agni re-indexes URL sources so the knowledge base reflects the current page content — useful for pricing pages and help-center articles that change over time.
How do I verify answers before going live?
Use the retrieval playground. Ask the questions your customers ask and inspect what the knowledge base returns. Combined with safe web-call testing in the agent builder, you can validate the full answer path before an agent speaks to a real customer.

More questions? Read the full FAQ or talk to sales.

Put your knowledge on the phone.

Bring a policy document or pricing page to the demo — we'll show you how an agent answers from it on a live call.

Talk to salesView API docs
or call our AI agent directly+91 11 4056 6601Available 24/7 · Hindi, Hinglish, EnglishChat on WhatsApp
Book a demo