Appearance
Balance & billing
The Sales Platform bills in AI messages. This page explains how that number is computed, what draws it down, what does not, what happens when it reaches zero, and how service resumes.
In the app
- The balance — "N Messages Left" — sits at the bottom of the sidebar with a Top up button, and in the Analytics header.
- Plans — app.mychatbot.app/subscription.
- Invoices and card — avatar → Settings → Billing → Manage on Stripe (/profile/payment). The button is disabled until the account has paid through Stripe at least once.
- Model and price tier — Settings → AI Core Settings (/profile/aiCore).
What an AI message is
- A message is a metering unit of AI work, not a chat bubble. One normal reply from a sales assistant costs one message, however long the conversation history it had to read.
- The number shown is weighted by your model tier. Models are priced per 100 messages, and the base tier is $4. On an $8 model the same underlying balance shows half as many messages; on a $14 model, 3.5× fewer.
| Model tier (per 100 messages) | 1 reply consumes | 600 base messages show as |
|---|---|---|
| $4 | 1 | 600 |
| $6 | 1.5 | 400 |
| $8 | 2 | 300 |
| $9 | 2.25 | 266 |
| $12 | 3 | 200 |
| $14 | 3.5 | 171 |
| $30 | 7.5 | 80 |
Changing the model on AI Core Settings therefore changes the displayed balance immediately — nothing was spent; the same funds are counted in bigger units. The tier list and prices in the app are authoritative; the Reasoning effort control appears only for the GPT-5 family (not the mini/nano variants).
- Voice plans show minutes: "N minutes left" is the same balance divided by five — 1 voice minute = 5 messages.
What is charged
| Action | Charge | Notes |
|---|---|---|
| A reply from a sales assistant on any channel | 1 message (×tier weight) | Includes replies composed for follow-ups and outreach |
| A voice call, per minute | Prorated by exact duration | Billed only when the call connects; a 3-second call costs a fraction of a message |
| Post-call functions (CRM actions after a call) | A flat $0.04 per call, converted to messages | Regardless of call length |
| Platform Wizard | 1 message per reply | The onboarding/configuration assistant |
| Sonar Pro Search (Web Search → Search model) | Extra message usage per search call | Plain Sonar is included |
Voice per-minute rate depends on the TTS model chosen on the Calls channel: $0.20/min for the v3 Conversational and Multilingual v2 models, $0.15/min for the Flash and Turbo family. That dollar figure is converted into your messages, so a minute costs the same number of dollars whatever text model you run. The Calls channel shows the result as "Rate: N AI messages per minute of call."
Additional call languages switch the voice tier
Turning on Additional languages for a Calls channel requires the v3 Conversational model — the app sets it for you, which moves the channel onto the $0.20/min rate.
What is free
| Action | Why it costs nothing |
|---|---|
| Test Dialog on the Agent configuration page | Rehearsing your own assistant is not billed. Tool runs from the editor's Test button are not billed either, but they execute — see Assistant configuration |
| The opening line of an outbound call | Charging it would double-bill answered calls and bill unanswered dials |
| AI client and conversation analytics | Recorded for cost visibility, never debited |
Because these are not billed, they are also absent from every Analytics number — the dashboard counts only billable usage.
Plans
Each Sales plan includes a monthly allowance of replies, sized at the $4 base tier:
| Plan | Included replies per month |
|---|---|
| Solo | 200 |
| Standard | 600 |
| Pro | 1 200 |
| Multipro | 3 000 |
| Corporate | 10 000 |
| Voice Builder / Launch / Scale / Corporate | 150 / 450 / 1 200 / 3 200 voice minutes (750 / 2 250 / 6 000 / 16 000 messages) |
Yearly plans credit twelve months at once. As the subscription page notes, "replies on premium AI models use proportionally more of your included messages."
Other plan rules that surprise people:
- Standard and Pro allow one active sales assistant. Every ten minutes the platform switches off all but the most recently updated assistant. Multipro and above have no such limit — see Assistant configuration.
- Trial: "Your 3-day free trial starts at checkout — no charge until it ends." End trial & pay now charges immediately. Cancelling means "You will lose access immediately."
- Contacts and indexed catalog capacity are separate meters with their own included amounts and overage prices — see Contact storage & pricing and Catalog capacity & pricing.
- The Agents Platform has its own dollar-denominated budget and top-ups ($5–$500) that never touch the Sales balance — see Agents usage & billing.
Topping up
Top up (sidebar) or Top up messages (Analytics) opens the AI Messages dialog: packages from 100 to 7 000 messages, priced at your tier ("Total cost for N AI Messages would be $X"), with Enable auto top-up and an "Auto top-up when below" threshold (default 200). Turning auto top-up off saves immediately; turning it on is saved with your next purchase. Accounts with a live Stripe subscription get a one-click confirmation; others go through checkout. Details and the voice-minute variant are on Analytics → Topping up messages.
When the balance runs out
Before zero, low-balance warnings go to the bell and by email at roughly 33, 17, 5 and 3 messages left (base tier), at most three times.
At zero:
- Every assistant that is on is switched off, with the reason recorded as insufficient tokens.
- One email per account is sent (not one per assistant), plus a Telegram note in each agent's Flight Control group: "I have been turned off due to insufficient tokens. Please top up your account."
- Chats keep arriving and stay visible; operators can still reply by hand. Customers get silence, not an away message. The chat header shows "Agent … is switched off, so nobody in this chat will get an automated reply. Turn it back on in AI configuration, or top up the balance if it ran out."
The same off-switch, with the reason subscription expired, fires when a subscription's end date passes without renewal. That also clears the plan, removes call webhooks, and starts a lapse clock: channels are disabled after 30 days and the account is marked for deletion after 60 days. Contact capture pauses with the channels.
How agents come back
Every payment that lands — a top-up, a renewal, a manual credit — re-enables the assistants that were switched off for a billing reason (insufficient tokens or subscription expired). An hourly backstop repeats the check for any funded account (a plan and a positive balance) whose muted assistants still have at least one live channel.
What is not re-enabled automatically:
| Assistant was switched off because… | After payment |
|---|---|
| You flipped its Status switch | Stays off. The owner's choice is never overridden — turn it on by hand |
| It was muted for billing before 2026-08-18 | Stays off (no reason was recorded then) — turn it on by hand |
| An instruction edit was flagged by moderation | Stays off — a suspension, not a fault |
| Plan agent limit (Standard/Pro second assistant) | Stays off until you upgrade or make it the most recently updated one |
| Working hours | Handled by the schedule itself; it flips back on when hours resume |
Two more rules: flipping the Status switch yourself — in either direction — clears any recorded reason, so from then on the platform treats the agent as owner-managed. And turning an assistant on without any plan is refused with "active subscription required to enable agent."
Check the Status switch after paying
If an assistant was off before the balance ran out, or was muted before mid-August 2026, paying will not wake it. Open Agents, pick the assistant, and confirm the switch reads "Agent … is working."
Use cases
- Budgeting a launch — 2 000 expected conversations at ~4 replies each on a $8 model ≈ 16 000 base messages: pick Multipro plus top-ups, and enable auto top-up so the assistant never stops mid-conversation.
- Cheap by day, smart by night — there is one model per account; if you need two tiers, run two accounts rather than expecting per-assistant pricing.
- Voice pilot — a Voice Builder plan is 150 minutes; at $0.15/min a 20-minute daily test burns it in a week — watch the minutes counter, not the dollar price.
Test it
- Note the sidebar balance, switch the model on AI Core Settings from a $4 to an $8 tier, and watch the number halve; switch back and it returns — nothing was spent.
- Send one message to your assistant on any channel: the balance drops by exactly the tier weight. Send one in the Test Dialog: it does not move.
- On a test account, let the balance reach zero: the assistant switches off, one email arrives, and the chat header banner appears. Top up 100 messages: within a minute the assistant is on again.
- Now switch an assistant off by hand, top up again, and confirm it stays off — that is the owner-managed rule.
See also
- Analytics & Quality Control → What the dollar costs mean — how the dashboard turns messages into dollars.
- Assistant configuration → Status — the switch itself, and the single-active-assistant rule.
- Contact storage & pricing — the contacts meter.
- Catalog capacity & pricing — the indexed-products meter.
- Calls (AI voice) → Billing — per-minute rates and post-call functions.
- Notifications — where the balance warnings arrive.
- Agents usage & billing — the separate Agents Platform budget.