Frequently asked questions about NeuronChat

Who benefits most:

NeuronChat is an AI agent and website chat widget platform. It fits teams that need a fast launch without an integrator and answers strictly from their own documents.

• —SMB with inbound from website and messengers
• —HoReCa, education, clinics/salons, retail, real estate, IT/digital
• —Sales and support teams with lots of FAQ and first-line qualification
• —Teams that need RU data handling and clear usage-based pricing

When NeuronChat may be overkill:

If you only need cold outreach at scale, a ticket helpdesk, or a complex WhatsApp BSP with broadcast to tens of thousands of contacts, look at specialized tools. NeuronChat is stronger at inbound: widget, company knowledge, qualification and CRM.

Out of the box:

Typical tasks the agent covers without code:

• —Answers about price list, services, terms and FAQ
• —Contact capture and interest qualification
• —Escalation to a live operator by rules
• —Leads into amoCRM, Bitrix24, Telegram alerts
• —Booking via integrations (YClients, calendar, iiko - on any paid plan)

How it learns:

You upload documents, FAQ or point to a website. The agent finds relevant fragments and answers from them. If nothing fits, it does not invent - it says so and offers to connect you.

The dashboard has unanswered questions: add the missing fact and the agent starts answering similar requests.

Quality depends on knowledge completeness: the more precise FAQ, price list and site pages are, the more confident the dialog. Expand the knowledge base from unanswered questions.

Plans count agents, and each agent works in exactly one channel: one website, one bot or one account. Start — 1 channel, Growth — 2, Business — 3, Pro — 5.

No need for two scenarios for website and Telegram: “Copy to another channel” creates the same agent with the same knowledge and rules — it takes one more plan slot.

When price list and FAQ change every week, redrawing bot-builder scenarios is usually the failure source. In NeuronChat you update a document or page and the agent uses current fragments.

For critical flows keep operator escalation and unanswered-question review.

Pros:

• —Widget live in minutes, no flowcharts
• —A trained agent copies to any channel in a minute
• —Fact control via knowledge base
• —Clear pricing: a plan with a token limit, usage visible in the dashboard

What to watch:

• —When the token limit is exhausted, visitors still get answers, but new dialogs and leads stay hidden in the dashboard until you buy a token pack or pay for a plan
• —Complex internal systems and SLA are easier on Enterprise
• —Omnichannel focuses on the RU stack (Telegram, WhatsApp, VK, CRM), not a global WhatsApp BSP suite

Typical path:

• —Sign up and create an agent
• —Upload FAQ/price list or crawl the site
• —Paste widget code (Tilda, WordPress, Bitrix, HTML and more)
• —Connect Telegram / CRM if needed

If you need help with an industry scenario, book a demo: sales will walk through rollout order.

• —Docs and widget install per platform
• —Free trial in the dashboard
• —Demo call for non-standard rollouts

Typical setup:

• —Widget / messenger → AI agent
• —Capture contact and request context
• —Create a lead or notify amoCRM / Bitrix24 / Telegram
• —Manager continues in the familiar CRM

What it sits next to:

• —Website: HTML widget (Tilda, WordPress, Bitrix and more)
• —Channels: Telegram, WhatsApp, VK
• —CRM: amoCRM, Bitrix24
• —Booking and services: YClients, calendar, iiko - per integrations catalog
• —In-chat payments: YooKassa

Each agent has its own knowledge base and its own channel (1 AI agent = 1 channel: one website, one bot or one account). That keeps tone and facts from mixing across brands. The Business plan includes 3 AI agents = 3 channels; beyond that, Pro or Enterprise.

Model:

Public plans are Start ($59/mo, 12.5M tokens (≈ 500 conversations)) and Business ($199/mo, 75M tokens (≈ 3,000 conversations)), with a discount for yearly billing. Conversation counts are estimates based on an average conversation, not a guarantee. Enterprise is custom: SLA, manager, custom integrations.

Signup includes 2.5M tokens (≈ 100 conversations) free for 30 days, no card required, so you can test the agent on your data.

If the limit is not enough you can buy a token pack — 280–480 ₽ per 1M tokens depending on the plan. A pack never expires and is spent after the plan limit.

What is usually outside the subscription:

Third-party channel costs (for example WhatsApp provider terms) and external integrator services are separate if you hire them. NeuronChat itself connects without a mandatory integrator.

AI models read and write text in pieces called tokens - a word or a part of one. NeuronChat counts usage in its own tokens: 1 token = 1 input token of the standard model. Agent replies and other models are converted into the same unit, so there is one limit whichever model answers.

Each plan includes a 30-day token limit: Start — 12.5M tokens (≈ 500 conversations); Growth — 30M tokens (≈ 1,200 conversations); Business — 75M tokens (≈ 3,000 conversations); Pro — 200M tokens (≈ 8,000 conversations). Conversation counts are estimates based on an average conversation, not a guarantee.

Why tokens rather than conversations: a two-line question and a forty-message conversation with product selection take very different amounts of work. With tokens you pay for the work the agent actually does - short conversations use the limit more slowly.

What affects usage:

• —Conversation length: every reply takes the conversation history into account, so long conversations use more
• —Knowledge-base context: the more fragments an answer needs, the more tokens
• —Model: economical models use the limit more slowly, premium models faster
• —Cache: repeated context (instructions, knowledge base) is counted at a lower rate

That is why conversation counts in plans are given as estimates: Start — 12.5M tokens (≈ 500 conversations); Growth — 30M tokens (≈ 1,200 conversations); Business — 75M tokens (≈ 3,000 conversations); Pro — 200M tokens (≈ 8,000 conversations). Actual token usage is visible in the dashboard.

This keeps budget under control: no money is charged for extra token usage without your decision. The continuation of conversations that were already running is hidden until payment too, while everything before that moment stays visible. If hidden conversations add up to half of the monthly plan limit, the agent stops answering in substance and asks visitors to leave their contact details - you get a separate notification about it. Hidden leads are not sent to your CRM or iiko until you pay (an order the customer already paid for also waits). For each new hidden event you get a short Telegram and email notice with its type only - "New lead", "New booking", "New paid order for N RUB" - plus a link to billing, grouped no more than once per 10 minutes. After you pay or buy a token pack, conversations and leads reopen under Dialogs and Leads, oldest first; each reopened conversation is charged the tokens it used while hidden. Free welcome tokens are not renewed. Daily caps can be configured separately in the dashboard if needed.

The plan token limit is granted for 30 days. When the next paid period starts the counter resets and you get the full limit again; the unused remainder of the previous period does not roll over.

Token packs are bought per 1M tokens - 280–480 ₽ per 1M tokens depending on the plan. They never expire: the plan limit is spent first and the pack only after it is exhausted. The remaining pack balance carries over to following periods.

The welcome 2.5M tokens are valid for 30 days after signup and are not renewed. If you need packs every month, moving to a higher plan is cheaper: tokens included in it cost less than packs.

• —Self-serve: widget, knowledge, core integrations, clear plan
• —Enterprise: manager, SLA, custom integrations and limits
• —Not sure - book a demo and we will suggest a path

The knowledge base, dialogs and leads are stored on servers in Russia, each company's data is isolated and dashboard access is role-based. Before launch read the privacy policy and public offer. Enterprise can include extended security and compliance terms on request.