Billing & plans

This page explains how Nerva charges for usage, what counts as a "minute" or "message", and what the plan tiers include. For current pricing in dollars, see the pricing page.

The two layers of cost

Two separate costs combine on your bill:

1. Your Nerva plan — the platform itself: dashboard, agent runtime, storage, analytics, post-call analysis.
2. Pass-through usage — what the underlying providers (LLM, STT, TTS, telephony) cost us, passed through to you. You only pay for what you actually use.

The Nerva plan is a fixed monthly fee (with included usage allowance). Pass-through usage is billed at the end of each cycle based on real consumption.

What counts as usage

Voice agents — voice seconds

Every second of call audio counts toward your plan's voiceSecondsPerMonth allowance:

• A 3-minute call = 180 voice seconds.
• A call that never connects (no-answer, busy, failed dial) = 0 voice seconds. You're never charged for unreachable contacts.
• Voicemail detection counts until the agent ends the call, typically 10–20 seconds.

Voice seconds bundle all the underlying costs (STT, LLM, TTS, telephony) into one number. Easy to predict.

Chat agents — chat messages

Every message exchanged (visitor in + agent out) counts toward chatMessagesPerMonth:

• A typical chat session of 10 back-and-forth turns = 20 messages.
• LLM token costs are bundled into the message price.

SMS / WhatsApp agents

Both inbound and outbound messages count toward chatMessagesPerMonth. Note that WhatsApp also charges Meta-side conversation fees on top — those are pass-through and billed at cost. See WhatsApp pricing.

Knowledge bases

• Indexing (one-time per document): counted against embeddingTokensPerMonth. A typical PDF burns 5K–50K tokens during indexing. Most SMBs never come close to the cap.
• Retrieval (during conversations): free — included in voice seconds and chat messages.

Recordings

• Recording storage counts against recordingStorageBytes. A 5-minute voice call generates ~3 MB of audio.
• Recordings older than your plan's retention window may be auto-archived or deleted (configurable per agent).

Campaigns

• Each campaign counts toward campaignsPerMonth.
• Each contact in a campaign counts toward maxCampaignContacts.
• Calls made during the campaign use voice seconds normally.

Plan tiers

We ship three plan tiers. Specific numbers may change — always confirm on the pricing page — but the shape is:

The Starter plan fits a single-location small business. Growth fits an agency or multi-location SMB. Enterprise is for high-volume operators with custom contracts.

Overages

If you exceed an allowance mid-cycle, three things can happen depending on which limit:

• Voice seconds / chat messages — overage billing kicks in. You pay per-second/per-message at a published rate above your allowance. No service interruption.
• Agent / phone number / KB count — creation is blocked until you upgrade or remove resources. Existing resources keep working.
• Concurrent calls — extra calls queue or are rejected with a busy signal, depending on your campaign settings.

Watch usage in real time on Dashboard → Settings → Usage.

Billing cycle

Plans renew on the same day each month (the day you signed up). At renewal:

• Plan fee is charged for the next month.
• Pass-through usage from the previous month is calculated and charged.

Both appear as separate line items on your Stripe invoice.

Switching plans

Dashboard → Settings → Billing → Change plan.

• Upgrading is instant. New allowances apply immediately; you're prorated for the remainder of the cycle.
• Downgrading takes effect at the next renewal. Resources that exceed the new plan's caps stay running until the renewal, then become read-only until you reduce.

Pause / cancel

• Pause an agent — toggle the agent to "Inactive". No new calls or chats route to it, but the configuration stays. No usage charges while inactive.
• Cancel the org — Settings → Billing → Cancel. Service continues until the end of the current billing cycle, then your org is deactivated. Data is retained for 30 days, then deleted.

Per-call cost visibility

Every call's detail page (Conversations → click any call) breaks down the cost:

• STT cost — speech-to-text provider
• LLM cost — language model provider
• TTS cost — text-to-speech provider
• Telephony cost — your phone provider (Twilio/Plivo)
• Total — the sum

These are the real provider costs, not Nerva's markup — useful for understanding which provider choices drive your costs.

Tips for SMB owners

• Start on Starter. 500 voice minutes is enough to validate the product with real customers. Upgrade once you're consistently bumping the cap.
• Watch concurrent calls during campaigns. If your plan allows 5 concurrent and you run a 100-contact campaign at concurrency=10, the campaign queues. Match concurrency to your plan.
• Cheap LLMs save real money. Switching a high-volume agent from GPT-4o to GPT-4o-mini can cut your per-call cost by 70–80% with minimal quality loss. Test it.
• Keep recordings reasonable. Long retention (years) adds up. Default 90-day retention is usually plenty.
• Use the KB strategically. Embedding token costs are tiny but cumulative — clean out old/duplicate sources you're not using.

See also

• Voice agents — voice-seconds pricing in context
• Knowledge bases — embedding token notes
• Campaigns — concurrent-calls limits during campaigns
• Phone numbers — provider charges (separate from Nerva)