Billing & Pricing

How usage is metered, what each meter costs, and how to read a charge back to the call that caused it.

How billing works

Converse is prepaid. You top up a wallet, and usage is drawn down against that balance as it happens. There is no monthly bill and nothing is invoiced in arrears — the only invoice you receive is a GST tax invoice covering usage you have already funded.

• Metered as it happens. The agent worker flushes usage every 30 seconds and once more at the end of the session, so a crashed worker loses at most half a minute.
• Idempotent. Every charge carries a key derived from the session and the tick number. A worker that retries after a network timeout bills once.
• Credit first, wallet second. Credit grants are drawn down oldest-expiring-first before the paid balance is touched.
• Tax held back. GST accrued on usage is subtracted from the spendable balance, so a call can never be funded with money already owed to the exchequer.
• In-flight calls finish. A session already running is allowed to take the balance negative; a new one is not started on empty.

Two billing products

Pay As You Go itemises agent orchestration, speech and exact language-model usage. Pay Per Second combines connected voice AI into a selectable Standard or Premium tier. Both use the same prepaid wallet, tax treatment, usage ledger and telephony meter.

Pay Per Second

Standard and Premium include speech, an eligible language model, agent runtime, tools and knowledge. Connected voice time is prorated by second with a ten-second minimum. Only models certified for the selected tier can be saved. Converse telephony remains ₹0.60 per started minute, producing the familiar ₹4 Standard or ₹6 Premium price for an ordinary full PSTN minute; web calls omit that phone-network line.

Pay As You Go

Each component follows its published billing unit. Agent orchestration and speech recognition are prorated by consumed second with a ten-second minimum per session. Language-model input and output use the selected model's token rate. Spectre Mini follows generated-audio duration; Spectre Pro follows generated characters. The final customer price is shown directly for every component.

Pay As You Go language-model prices

The selected model's current input and output rates are resolved once for the session and written to the immutable usage event. With a workspace-owned provider key, that provider bills the workspace directly and Converse does not duplicate the speech or model charge.

Which models your plan can reach

A model is selectable when it is healthy, available in your region and has either a platform rate or a valid workspace-owned provider key. Pay Per Second further limits selection to models whose cost fits Standard or Premium. The builder shows an unavailable model with its reason instead of silently substituting another bill.

The decision is enforced server-side as well as in the agent form, so it holds no matter which supported client creates or updates an agent.

Agent Builder is the authoritative model catalogue for a workspace. It shows both usable and unavailable choices with a customer-safe reason; the public SDK does not currently expose a model-catalogue endpoint.

Model lists are never hardcoded — each provider is asked for its own current catalogue, so a model is available here the day the vendor ships it. The Others group is everything reachable through OpenRouter, minus anything Google, OpenAI or Anthropic sell directly, since those have their own sections and listing them twice would offer the same model on two different bills.

Telephony

Converse telephony costs ₹0.60 for each started minute on every answered PSTN leg, including an answered transfer or conference leg. Browser and in-app sessions do not use this meter. A workspace using its own carrier pays that carrier directly.

Channels

WhatsApp is priced by Meta's own conversation categories, because that is how Meta bills. Service replies inside a user-initiated 24-hour window carry no Meta charge and are free here too. You supply your own Meta WhatsApp Business account.

Platform

Selectable model rates

The public catalogue lists the final customer rates for supported selectable models. It lets you size platform-funded or bring-your-own-key usage before committing without exposing Converse's internal service implementation.

curl https://api.converse.axllabs.in/v2/pricing/providers

{
  "fx": { "usd_inr": 95.24, "as_of": "2026-08-10" },
  "providers": [
    {
      "key": "llm.google.gemini_25_flash",
      "provider": "Google",
      "product": "Gemini 2.5 Flash",
      "service": "llm",
      "unit": "1,000 tokens",
      "rate_in_inr": 0.02857,
      "rate_out_inr": 0.2381,
      "byok": true
    }
  ]
}

Rates carrying rate_in_inr and rate_out_inr are language models, priced per direction. Everything else uses the single rate_inr field. fx.usd_inr is the rate every USD-denominated figure was converted at, so you can tell a repricing from a rupee movement.

The pricing API

Unauthenticated — it is the price list on the website. Nothing organisation-specific is served here.

• GET /v2/pricing — the whole catalogue: meters, plans, provider rates, GST rate and FX basis.
• GET /v2/pricing/plans — plans only, with rates fully resolved.
• GET /v2/pricing/providers — selectable model and channel rates.
• GET /v2/pricing/quote — what a workload would cost.

Responses carry an ETag derived from the catalogue version and are cached for five minutes by the Converse API, so a page that reloads on every navigation normally receives a lightweight 304 response.

Quoting a workload

curl "https://api.converse.axllabs.in/v2/pricing/quote\
?plan=free&stack=standard&minutes=10000&telephony=true"

• plan — free for Pay As You Go, per_second for fixed voice, or another active plan key.
• stack — on Pay Per Second, standard selects ₹4 PSTN voice and premium selects ₹6. On Pay As You Go it selects the representative speech stack used by the estimate.
• minutes — connected minutes per month. Defaults to 1,000.
• telephony — false for web and in-app sessions, which have no carrier leg.

The response itemises each meter, a pre-tax subtotal and an effective pre-tax per-minute figure. Pay Per Second quotes the selected fixed voice meter directly. Pay As You Go adds its component assumptions: roughly 35 generated speech seconds (or 850 characters), 8,000 input tokens and 350 output tokens per connected minute. Your own agents will differ; the response says so.

Reading a charge back to a call

Every charge creates immutable billing evidence carrying the meter, quantity, rate in force, allowance applied, and the call that caused it. Existing evidence is never rewritten when a rate changes, so historical invoices remain reproducible.