Frequently asked
questions.

Agni is a real-time voice AI platform built specifically for Indian businesses. It supports 11 Indian languages including Hindi, Hinglish, Tamil, Telugu, Kannada, Marathi, Bengali, Gujarati, Punjabi, Malayalam, and English — with sub-300ms latency. It handles outbound campaigns, inbound routing, IVR flows, appointment booking, and post-call analytics out of the box.

Global platforms fail on Hinglish code-switching, Indian accents, and regional dialects. Agni is built ground-up for Bharat — with Thunder Emotion AI model tuned on Indian conversational data, 28 voices with native-sounding accents, RBI/DPDP compliance built in, and INR pricing with no dollar conversion surprises.

Agni uses the Thunder Emotion model — a voice AI provider optimised for Indian conversational patterns. It includes temperature control for tuning response style, and supports advanced settings for fine-tuning model behaviour per agent.

Agni achieves sub-300ms voice latency for most calls. This is fast enough that callers cannot tell they're speaking to an AI — critical for collections, telesales, and any situation where a pause breaks rapport.

No. You can sign up and explore the full Agni platform for free without entering a credit card. Paid plans start at ₹2,999/month.

Agni supports 11 languages natively: Hindi, Hinglish (code-switching), Tamil, Telugu, Kannada, Marathi, Bengali, Gujarati, Punjabi, Malayalam, and Indian English. Starter plan includes Hindi + English; all other plans unlock all languages.

Hinglish — the natural code-switching between Hindi and English spoken by 350M+ urban Indians — is Agni's primary strength. The model handles mid-sentence language switching, Romanised Hindi input, and urban slang without missing a beat. This is the use case global AI platforms fail at most.

Agni offers 28 voices including Iris, Brennan, Rowan, Celeste, Theodore, Finn, Vivienne, Patrick, Stella, Luna, Dominic, Julian, and more — across male and female options with varying regional accents. Each agent can be configured with a different voice.

Yes. Agni detects the language the caller is responding in and adapts mid-call. This is particularly effective for Hindi/Hinglish/English switching, which happens naturally in almost every Indian business call.

Each agent supports: End Call (graceful termination), Transfer Call — Static (to a fixed number) or Dynamic (to a number determined at runtime), IVR / Press Digit flows (listen for DTMF input and branch), and Custom Functions (execute a webhook/API call mid-call and use the response to continue the conversation).

Static transfer routes the call to a predefined phone number when a condition is met (e.g., customer wants to speak to a human). Dynamic transfer uses a Custom Function to fetch the right number at runtime — useful for round-robin agent queues or territory-based routing.

Custom Functions let Agni make a live API/webhook call during a conversation — for example, to look up a customer record, check loan status, verify OTP, or book an appointment. The agent receives the API response and continues the conversation naturally based on the data returned.

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.

Temperature controls how creative vs. deterministic the AI's responses are. Lower temperature = more consistent, scripted responses (better for collections and compliance). Higher temperature = more natural, adaptive conversation (better for telesales and support). You can tune this per agent in AI Provider settings.

Per agent you can set: Background Sound (add ambient audio like a call centre environment), Interruption Sensitivity (0–1 slider controlling how easily the caller can interrupt the agent mid-sentence), and Reminder Message Frequency (how often the agent re-prompts an unresponsive caller).

Agni detects when a call hits voicemail rather than a live answer. When detected, it plays a pre-recorded voicemail message (you set this) and ends the call. Detection timeout is configurable (default 10 seconds). This prevents the AI from treating voicemail greetings as live conversation.

"End call on silence" is configurable — by default the agent waits 10 seconds of silence before ending the call. You can increase or decrease this threshold based on your use case.

Yes. Max call duration is configurable up to 30 minutes per call. Ring duration (how long the phone rings before giving up) is also configurable, with a default of 30 seconds.

Memory allows the agent to retain context across multiple calls with the same contact. When enabled, the agent remembers previous conversation details — useful for follow-up calls, collections reminder sequences, or multi-step onboarding flows. Available on Scale and Enterprise plans. Memory carries a metered charge of ₹2/min when active during a call.

After every call, Agni automatically extracts: Sentiment classification (Positive / Neutral / Negative), Call summary (generated by GPT-4o mini — a concise paragraph of what was discussed and the outcome), and any structured data fields you define in the agent schema.

The Analytics dashboard shows: Usage trend (calls over time), Call outcomes (completed / voicemail / failed / no answer), Sentiment analysis breakdown across all calls, and Agent performance metrics. Filters are available by agent, campaign, and date range.

Yes. Call Sessions can be filtered by Web (demo calls), Inbound, and Outbound. Each session shows the sentiment score, duration, outcome, and a link to the transcript and recording.

Call recording is available on Enterprise plans and stored on India-only servers for up to 2 years — meeting RBI requirements for NBFC and collections use cases. Scale and lower plans include transcripts and summaries but not raw audio recordings by default.

Campaigns let you run bulk outbound calling at scale. You upload a contact list, assign an agent, configure call scheduling, and Agni dials out automatically. Campaigns can be in Running, Scheduled, Paused, or Draft state. You can run multiple campaigns simultaneously.

Contacts can be uploaded via CSV import or added individually. Each contact has tags, status (Fresh Lead / Need Follow-up / Failed — Retry), and full call history attached. You can export contacts with outcomes for CRM sync.

Yes. Inbound calling assigns a phone number to an agent — when that number receives a call, Agni answers with the configured agent. You can run multiple inbound agents on different numbers simultaneously.

For collections use cases, Agni enforces: calls only between 8AM–7PM, maximum 2 calls per day per number, consent disclosure at the start of every call, and full call recording for audit. These are hard-coded compliance guardrails, not optional settings.

Live integrations include: Twilio (telephony), Airtel India, Telynex, GoHighLevel / Lead Connector (CRM + calendar), Cal.com (appointment booking), Salesforce, HubSpot. Coming soon: Zoho CRM, Freshdesk, Leadsquared. Any system can be connected via REST API or webhooks.

The GoHighLevel integration is a calendar integration — Agni books appointments directly into GHL during live calls. It checks calendar availability in real time, confirms slots with the caller, and writes the booking back to GHL. Pipeline triggers and CRM sync are also supported.

Same as GoHighLevel — Agni checks Cal.com availability live during a call, confirms with the caller, and books the appointment. The caller receives a calendar invite automatically. Rescheduling is also handled in-call.

You can configure a webhook URL per agent to receive real-time event notifications — call started, call ended, voicemail detected, transfer triggered, custom function executed, etc. This is how you pipe Agni data into your own systems without using a native integration.

REST API access and API key management are available from the Scale plan (₹12,999/month) and above. Starter and Growth plans use the Agni dashboard and native integrations, but do not include direct API access.

Agni supports Twilio phone numbers and custom provider SIP trunking. Enterprise customers can connect via SIP/TLS, RTP/SRTP, WebRTC, or direct PSTN interconnect through Indian carrier partners (Airtel, Telynex, and others).

Yes. Agni is designed for India's Digital Personal Data Protection Act. Data is stored on India-only infrastructure, caller consent is captured and logged, data retention periods are configurable, and access controls meet DPDP requirements. DPDP compliance suite is included from the Scale plan.

Yes. RBI's Fair Practice Code compliance is built into the collections workflow — 8AM–7PM call windows enforced at the system level, 2-calls-per-day limit per number, consent-first call opening, and 2-year call recording on Enterprise. These are not optional settings.

All call data, recordings, transcripts, and customer information is stored on India-only servers. No data is routed through or stored in US/EU infrastructure. This satisfies both DPDP and RBI data localisation requirements.

Yes. The Agni platform supports 2FA for all user accounts, configurable under Security settings. Enterprise organisations can also enforce SSO via your own identity provider.

SLA (Service Level Agreement) with uptime guarantees is available on Enterprise plans. This includes a committed uptime percentage, incident response time targets, and a dedicated support contact. Scale and below operate on best-effort availability without a formal SLA.

Agni pricing has two components: a monthly base fee (₹2,999 / ₹5,999 / ₹12,999 / custom) which includes a bundle of free minutes (300 / 1,000 / 2,500), and an overage rate for minutes beyond that bundle (₹9.5 / ₹8.75 / ₹8 / ₹7 on Enterprise). All prices are excl. 18% GST.

Starter (₹2,999/mo — 300 mins included, ₹9.5/min overage): 1 AI agent, 5 concurrent calls, 2 knowledge bases, Hindi + English, 28 voices with Thunder Emotion model, voicemail detection, post-call summary and sentiment, webhook notifications, and basic analytics. Email support included.

Growth (₹5,999/mo — 1,000 mins included, ₹8.75/min overage): 10 AI agents, 15 concurrent calls, 5 knowledge bases, all 10+ Indian languages, Hinglish + dialect fine-tuning, call transfer (static + dynamic), IVR/press-digit flows, custom webhook functions, GoHighLevel + Cal.com calendar integrations, outbound campaigns, contact management with CSV import, advanced analytics, call recordings, and priority support.

Scale (₹12,999/mo — 2,500 mins included, ₹8/min overage): Everything in Growth plus 25 concurrent calls, up to 15 knowledge bases, full REST API access, API key management, Twilio + custom phone providers, inbound call routing, agent memory (cross-call context), and DPDP compliance suite. Dedicated account manager included. API and integrations access starts at this plan.

Enterprise (custom pricing ₹7/min): Everything in Scale plus 10,000+ concurrent calls, unlimited API rate limits, SLA guarantee, 2-year call recording, on-prem/private cloud deployment, dedicated compliance team, custom CRM + CBS integrations, SOC 2 / HIPAA readiness, multi-org user management, and 24/7 dedicated support.

The Scale plan supports up to 25 concurrent calls. If your business needs more simultaneous calls, Enterprise is the right fit — with 10,000+ concurrency and custom pricing. Contact info@ravan.ai to discuss.

Yes. Annual billing comes with significant discounts. Contact info@ravan.ai or reach out via the Agni dashboard to discuss annual pricing.

Yes. Three features carry metered charges on top of your base plan: (1) Extra knowledge bases beyond your plan allowance (2 on Starter, 5 on Growth, 15 on Scale) cost ₹799/KB/month each. (2) KB live retrieval — when your agent queries a knowledge base during a live call — costs ₹1/min on top of your base per-minute rate. (3) Agent Memory, when active during a call, costs ₹2/min extra. These charges only apply when the features are in active use.

KB live retrieval is when your agent searches a knowledge base in real time during a phone call — for example, to answer a product question or look up a policy document. You are charged ₹1/min for minutes in which the agent performs KB lookups. If a call does not involve any KB query, this charge does not apply. The charge stacks on top of your normal per-minute platform rate.

Agent Memory allows your agent to remember context from previous calls with the same contact — so it knows what was discussed last time without the caller repeating themselves. When Memory is enabled on an agent and that agent is on an active call, ₹2/min is charged on top of your base per-minute rate. Memory is available on Scale and Enterprise plans and can be toggled per agent in settings.