Skip to content
codebrix

Cost guide

AI chatbot development cost: build price and monthly running cost

At Codebrix, building an AI chatbot into your website or product costs an indicative ₹60,000 to ₹2,00,000 ($2,200 to $7,500) and takes 2 to 4 weeks. After launch you pay running costs: LLM API usage billed per token, hosting, and Meta's per-message charges if the bot runs on WhatsApp. This guide prices the build and shows a formula for the monthly bill.

Build
₹60,000–2,00,000 indicative
Timeline
2–4 weeks
Running cost
Per token, plus channel fees
Models
OpenAI or Anthropic

Updated · By the Codebrix team

What you are paying for in a chatbot build

A useful AI chatbot is not a prompt pasted into a widget. It answers from your own information, knows when it does not know, hands over to a person, and can be checked before customers see it. The build cost covers these parts.

  • Conversation design: what the bot handles, what it refuses, and the tone it uses.
  • Knowledge base: loading your FAQs, documents, product data or policies, usually into a vector database so the bot retrieves the right passage before it answers.
  • Actions: looking up an order, checking a slot, creating a lead in your CRM. Each one is an integration.
  • Handoff: passing the conversation to a human by email, chat inbox or WhatsApp, with the transcript.
  • Guardrails and testing: a set of real questions we run against every change, so answers do not drift.
  • Admin and analytics: conversation logs, unanswered questions and cost tracking.
  • Channels: website widget, in-app chat, WhatsApp or more than one.
Build cost drivers
DriverKeeps cost lowPushes cost upEffect
Knowledge sourcesOne set of FAQs or documentsSeveral systems kept in syncHigh
ActionsAnswers questions onlyBookings, order lookups, CRM updatesHigh
ChannelsWebsite widgetWebsite, app and WhatsAppMedium
LanguagesOne languageSeveral, including mixed-language inputMedium
Testing depthSpot checksA scored question set run on every changeMedium
Data sensitivityPublic information onlyPersonal or account data, access rulesHigh

Codebrix indicative build ranges

Our AI assistant and automation work is an indicative ₹60,000 to ₹2,00,000 ($2,200 to $7,500) and takes 2 to 4 weeks when added to a site or product. The lower end covers a website assistant answering from one knowledge source, in one language, with handoff to email. The upper end covers assistants that take actions through your systems, pull from several sources, run on WhatsApp as well as the web, and ship with a scored test set.

If you do not have a site or app for the chatbot to live in, that build is priced separately: a marketing website is an indicative ₹40,000 to ₹1,20,000 and a web application ₹1,50,000 to ₹6,00,000. All of these are indicative; the fixed price comes after a discovery call. Our AI chatbot development page describes the build itself, and AI integration covers assistants inside an existing product.

Running costs: how LLM token pricing works

Model providers bill per token, a chunk of text roughly the size of a short word. You pay for input tokens (everything sent to the model: instructions, retrieved documents, conversation history and the user's message) and output tokens (the reply). Output costs more per token, but input is usually the bigger volume, because every turn resends the instructions and context.

Selected API prices per 1 million tokens, standard tier, as of November 2026
ModelInputOutputCached input or cache hits
OpenAI gpt-6.1-terra$1.00$5.00$0.05
OpenAI gpt-6-luna$0.10$0.50$0.01
Anthropic Claude Sonnet 5.5$2.00$10.00$0.10
Anthropic Claude Haiku 5.5 (prompts up to 100,000 tokens)$0.10$0.50$0.01

These are a sample, not the full price lists, and providers change prices and models often. Check the OpenAI and Anthropic pricing pages linked at the end of this guide before you budget. Anthropic also notes that its newer Claude models use a tokenizer that produces roughly 30% more tokens for the same text, so a token count measured on one provider does not transfer exactly to another.

A worked monthly estimate

Monthly token cost = conversations per month × turns per conversation × (input tokens per turn × input price + output tokens per turn × output price) ÷ 1,000,000.

Assumptions for this example, stated so you can swap in your own: 3,000 conversations a month; 6 turns each; 2,000 input tokens per turn (about 800 of instructions, 900 of retrieved documents and 300 of history and the question); 250 output tokens per reply; no caching. That gives 36 million input tokens and 4.5 million output tokens a month.

Example monthly token cost (assumptions above, prices as of November 2026, before tax)
ModelInput costOutput costMonthly total
OpenAI gpt-6.1-terra36 × $1.00 = $36.004.5 × $5.00 = $22.50$58.50
Anthropic Claude Sonnet 5.536 × $2.00 = $72.004.5 × $10.00 = $45.00$117.00
OpenAI gpt-6-luna36 × $0.10 = $3.604.5 × $0.50 = $2.25$5.85
Anthropic Claude Haiku 5.536 × $0.10 = $3.604.5 × $0.50 = $2.25$5.85

WhatsApp channel costs

Running the same assistant on WhatsApp adds Meta's charges on top of token costs. As of November 2026, Meta prices the WhatsApp Business Platform per delivered message, a model that replaced conversation-based pricing on 1 July 2025. Rates depend on the message category (marketing, utility, authentication, service) and the recipient's country, and you are charged only when a message is delivered.

  • Service messages, the replies a support bot sends, became chargeable again from 1 October 2026, with 1,000 free service messages per month on each business phone number, resetting monthly.
  • Messages sent inside a free entry point window opened by a Click to WhatsApp ad are not charged, with some exceptions Meta lists.
  • Meta can update rate cards only on set quarterly dates, so check the rate card for your target countries before launch.
  • If you use a WhatsApp Business Solution Provider, it may add its own fee on top of Meta's.

Because these charges scale with every reply, we design WhatsApp bots to resolve in fewer messages and to hand over to a person early when the bot is not helping. See WhatsApp chatbot development for how we build them.

Other recurring costs and hidden costs

Costs beyond tokens
ItemTypeWhat to know
Hosting for the bot's backendMonthlySmall at low volume; grows with traffic
Vector databaseFree tier or monthlyDepends on document volume and provider
Embedding the knowledge basePer token, mostly one-timeRe-run when documents change
Monitoring and logsMonthlyNeeded to spot wrong answers and cost spikes
Content upkeepYour team's timeOut-of-date documents produce out-of-date answers
Model changesOccasionalProviders retire models; switching needs a test run
Support retainerOptional monthlyIndicative ₹15,000–50,000 per month with us

The hidden cost people miss most is evaluation. A bot that answers 90 out of 100 questions well will still upset the ten customers it gets wrong. Budget time to review conversation logs every week for the first month.

How to cut running cost without hurting answers

  1. Route by difficulty. Send simple questions to a small model and only hard ones to a larger model.
  2. Trim context. Retrieve the three best passages, not twenty. Fewer input tokens per turn is the biggest lever.
  3. Use prompt caching for long, repeated instructions where the provider supports it.
  4. Cap history. Summarise older turns instead of resending the whole conversation.
  5. Set per-user rate limits and a monthly spend alert with the provider.
  6. Answer the top questions with fixed responses or buttons when a model adds nothing.
  7. Ask whether you need a chatbot at all. A clear FAQ page or a better search box is sometimes the cheaper fix, and we will say so.

How to get an accurate chatbot quote

Send 30 to 50 real questions your customers or staff ask, the documents that answer them, the channels you want, the systems the bot must read from or write to, and your expected monthly volume. With that, we can scope the build and project the monthly running cost on the 30-minute discovery call, then send a fixed-price proposal with milestones within 2 to 4 days. API keys and accounts sit in your name, so the token bill goes straight to you with no markup. If you want the bot to run whole workflows rather than answer questions, look at AI automation, or ask us for a fixed quote.

Frequently asked questions

How much does it cost to build an AI chatbot?

With Codebrix, an AI assistant built into an existing site or product is an indicative ₹60,000 to ₹2,00,000 ($2,200 to $7,500), taking 2 to 4 weeks. A simple bot answering from one knowledge source sits at the lower end. Bots that take actions, use several sources or run on WhatsApp sit higher. The fixed price follows a discovery call.

What does an AI chatbot cost to run each month?

Mostly LLM tokens, plus hosting, a vector database and channel fees. Using our worked example of 3,000 conversations of six turns, the token bill came to $58.50 on OpenAI gpt-6.1-terra and $117 on Claude Sonnet 5.5, at November 2026 prices. Smaller models cost far less. Your real figure depends on volume and context size.

Is OpenAI or Anthropic cheaper for a chatbot?

It depends on the models you compare and how your prompts are built. Both providers offer large and small models at very different prices, and their tokenizers count text differently. We test your real questions on two or three candidate models during the build and choose on answer quality and measured cost, not list price alone.

Does a WhatsApp chatbot cost more than a website chatbot?

Usually, yes. On top of token costs, Meta charges per delivered message by category and country, as of November 2026. Service messages became chargeable again from 1 October 2026, with 1,000 free service messages a month on each business phone number. A Business Solution Provider may add its own fee. Check Meta's rate card for your markets.

Who pays the OpenAI or Anthropic bill?

You do, directly. We create the API account in your company's name, or use yours, so the provider bills you with no markup from us. We set spend alerts and rate limits so a spike in usage or misuse cannot run up an unexpected bill.

Can the chatbot answer from our own documents?

Yes. We load your documents, FAQs or product data into a searchable store, and the bot retrieves the relevant passage before answering. It can show where an answer came from and say it does not know when nothing relevant is found. Keeping those documents current is the main ongoing task for your team.

Sources

  • AI chatbots

    An AI chatbot development company for assistants grounded in your content

    AI chatbot development company in Noida. Assistants that answer from your own documents, hand off to people and are tested before they go live.

    Read more
  • WhatsApp automation

    WhatsApp chatbot development on the official Business Platform

    WhatsApp chatbot development on Meta's Cloud API: approved templates, opt-in, handoff to your team and Meta's per-message pricing explained upfront.

    Read more
  • AI integration

    AI integration services for the software you already run

    AI integration services that add LLM assistants, document search, summaries and classification to your product, tested on your data with cost limits set.

    Read more
  • AI automation

    An AI automation agency for the repetitive work behind your business

    AI automation agency that moves email triage, lead follow-up, CRM updates and WhatsApp replies to AI workflows, with people approving the risky steps.

    Read more
  • Hire developers

    Hire AI developers who build, test and ship LLM features

    Hire AI developers who build LLM features with retrieval, agents and evals in Python or TypeScript. Paid first milestone, your keys, rate after a call.

    Read more
  • Cost guide

    How much does website development cost in India?

    Website development cost in India, explained: what moves the price, our indicative ranges, recurring costs, hidden costs and how to get a fixed quote.

    Read more
Estimate Call WhatsApp