Skip to content

Claude for your website

A Claude chatbot for your website, trained on your own content

Pick Claude Haiku 4.5, Claude Sonnet 5.5 or Claude Opus 5.5 as your agent's model, and it answers visitors from your website, files and Q&A, with the same actions, inbox and channels as any intoCHAT agent. You don't need an Anthropic account or API key.

Example: Claude Sonnet 5.5 answering from a shop's own returns page

Visitor: I bought a jacket three weeks ago and it doesn't fit. Can I still return it?

Agent: Yes, you are still in time: unworn items can be returned within 30 days of delivery. Start the return from your order confirmation email and print the free label. The refund goes to your original payment method within 5 working days of the jacket reaching our warehouse.

Visitor: Could I swap it for a larger size instead?

Agent: We don't do direct exchanges, but you can return this jacket and order the larger size at the same time, so you don't have to wait for the refund first.

Claude on its own does not know your business

Claude is a capable model, but it has never read your returns policy, your price list or this week's opening hours. Building a support bot on the raw API means writing the knowledge search, a chat widget, lead forms, a hand-off to your team and an inbox yourself. intoCHAT is that layer. It trains an agent on your content, lets the Claude model you choose write the replies, and gives you the widget, the actions and the inbox around it.

What you get with Claude on intoCHAT

Anthropic's models write the replies; intoCHAT supplies your knowledge, the tools and the place your team works.

  • Three Claude models to choose from

    Claude Haiku 4.5 for quick answers at 1 message per reply, from Starter. Claude Sonnet 5.5 at 2 messages per reply and Claude Opus 5.5 at 5, from Pro.

  • Answers from your own knowledge

    Claude writes each reply from the excerpts intoCHAT finds in your website pages, files, text and Q&A pairs, Notion pages or Google Drive files, and follows your instructions and guardrails.

  • The same actions as every agent

    API actions, client-side actions, chat forms, procedures, booking, product search and hand-off to your team work with Claude just as they do with the default model.

  • The same inbox and channels

    Claude writes the replies in the chat widget, the inline iframe and chat link, the Help Page, Slack and email. Your team reads every conversation and takes over from the same Inbox.

  • A fallback when Anthropic is slow or down

    If Anthropic fails, or takes more than 25 seconds before the reply starts, GPT-6 Luna answers instead and the reply uses one message.

How to set up a Claude chatbot

  1. Create and train your agent

    Create an agent, add your website, files or Q&A pairs, and train it. Write instructions for tone, scope and what to say when the answer isn't in your knowledge.

  2. Choose a Claude model

    On the Setup tab, open the Model card and pick Claude Haiku 4.5, Claude Sonnet 5.5 or Claude Opus 5.5. Try it in the Playground before you save, or put it next to GPT-6 Luna in Compare.

  3. Put it where your customers are

    Add the script tag, the WordPress plugin or the Shopify snippet to your site, or share your Help Page. From Starter you can also connect Slack or give the agent its own email address.

Claude models on intoCHAT

  • Claude Haiku 4.5

    Fast tier, 1 message per reply, on Starter, Pro and Enterprise. The quickest replies, for simple questions at high volume. It is also one of the models Auto can pick.

  • Claude Sonnet 5.5

    Advanced tier, 2 messages per reply, on Pro and Enterprise. Careful, well-written answers and reliable tool use.

  • Claude Opus 5.5

    Premium tier, 5 messages per reply, on Pro and Enterprise. The most capable, for complex questions and multi-step actions.

  • Images yes, PDFs no

    Claude reads images that visitors attach. PDFs are read by OpenAI's models only: with Claude, the agent is told a PDF was attached and answers from what the visitor wrote.

  • See which model answered

    Why this answer under each reply names the model that wrote it, and the Models used card on the Dashboard counts replies per model.

  • Mix models across agents

    Each agent has its own model setting, a fixed model or Auto. Run Claude Sonnet 5.5 on a sales agent and GPT-6 Luna on a busy FAQ agent in the same account.

Frequently asked questions

Can I build a website chatbot with Claude?

Yes. Create an intoCHAT agent, train it on your content, and pick Claude Haiku 4.5 (from Starter) or Claude Sonnet 5.5 or Claude Opus 5.5 (from Pro) as its model. Claude then writes the replies visitors get in the chat widget on your site. You don't need an Anthropic account, an API key or any code beyond the embed snippet.

Is intoCHAT made by Anthropic or partnered with it?

No. intoCHAT is an independent product. It uses Anthropic's Claude models through Anthropic's API, in the same way it uses OpenAI's and Google's models through theirs. intoCHAT isn't affiliated with or endorsed by Anthropic. Claude and Anthropic are names belonging to Anthropic.

Which Claude model should I use for customer support?

Start with Claude Haiku 4.5 for many simple questions: it gives the quickest replies and uses one message per reply, the same as the default model. Move to Claude Sonnet 5.5 when answers need more care or the agent runs several actions in a conversation. Keep Claude Opus 5.5 for complex questions and multi-step actions, since each reply uses 5 messages. Compare them side by side in the Playground before you save.

Does Claude work with actions, forms and hand-off?

Yes. Your agent's knowledge, instructions, guardrails, actions, forms and procedures stay the same with any model. Larger models usually take longer to answer than GPT-6 Luna and the fast models, especially in replies that run actions.

What happens if Anthropic's API is down?

If Anthropic fails before the reply starts, or doesn't respond within 25 seconds, intoCHAT tries GPT-6 Luna once instead. That reply uses one message: the extra messages of Sonnet or Opus are given back. A reply that has already started streaming can't switch models, so it may stop short in rare cases.

Where does visitor data go when Claude answers?

The visitor's messages, the conversation so far, your instructions and the knowledge excerpts used for the reply go to Anthropic. intoCHAT's background work, such as searching your knowledge and spam checks, always uses OpenAI, so messages also reach OpenAI. If your privacy policy names the companies that process chat messages, list both.

Can I use Claude on the Free plan?

No. Agents on Free answer with GPT-6 Luna from OpenAI. Claude Haiku 4.5 is included from Starter, and Claude Sonnet 5.5 and Claude Opus 5.5 from Pro. Starter starts with a 7-day free trial.

Put Claude to work on your own content

Create your agent free, train it on your website, then switch it to a Claude model on Starter or Pro and try it in the Playground.

Create your agent free

Free plan, no credit card needed.