Question 1

Which LLM's models are supported?

Accepted Answer

By default our bots use GPT-4o Mini Standard, Pro & Enterprise tier subscribers can use advanced LLM models such as GPT-5 from a wide variety of providers including OpenAI, Google, Cohere, Claude & Mistral. See available models for a full list of supported models.

Question 2

I will need more storage/message tokens than your plans allow, what should I do?

Accepted Answer

If your needs don’t fit into any of our standard plans, we can happily make your a custom plan. Please email hello@chatthing.ai with your requirements.

Question 3

Can I see the questions people are asking my bots?

Accepted Answer

Yes you can! Check out the conversations view after logging into your account.

Question 4

What is a data source?

Accepted Answer

Data sources are the way you connect you bot to your data. Currently we support importing data from Notion, Web pages a selection of text file formats & youtube videos but plan to add more very soon.

Question 5

What is a storage token?

Accepted Answer

Storage tokens are used when you synchronise your data from a data source such as Notion with your bot. 1000 storage tokens roughly equals about 750 words. Each time you synchronise a data source it will use more storage tokens.

Question 6

What is a message token?

Accepted Answer

Message tokens are used each time you send a message to a bot and receive a response. Each message uses between 1000-4000 message tokens depending on the questions and the bots response. This means 200,000 message tokens roughly equates to 100 messages.

Question 7

Can I use the Chat Thing API?

Accepted Answer

Yes, you can! We've just released our public API which is available to users on the Standard, Pro & Enterprise tiers. See the full documentation, for more information.

Question 8

Where can I get help?

Accepted Answer

You can get help by visiting our documentation or by joining our discord community

Question 9

How does our Chat Thing use Google user data?

Accepted Answer

Chat Thing can make use of files stored within your Google Drive. You have full control of which files or folders you grant us access to & any such information will only be used to train your bots. Any use of information received from Google APIs adheres to Google API Services User Data Policy, including the Limited Use requirements.

Question 10

Where can I find my previous invoices?

Accepted Answer

You can access your previous invoices by visiting https://chatthing.ai/app/billing and clicking on the "Manage subscription" button in the top right of the screen.

Question 11

If you're having trouble logging in?

Accepted Answer

Forgotten your password? You can reset it here. You may also need to confirm your email address, if you can't find this email please check your junk folder.

Question 12

Roughly how many messages do I get on each plan?

Accepted Answer

Message tokens are easier to think about as conversations. A typical customer question and bot answer uses 1,000–4,000 message tokens — longer conversations use more, because the chat history is sent with each message. As a rough guide: Plan — Message tokens / month — Roughly Free — 100,000 — 25–100 messages Basic — 3,500,000 — 900–3,500 messages Standard — 19,000,000 — 4,750–19,000 messages Pro — 38,000,000 — 9,500–38,000 messages Enterprise — 120,000,000 — 30,000–120,000 messages The simplest way to think about the two token types: storage tokens are what you put in (syncing your website, docs and files), message tokens are what your customers use up (chatting with your bot). Allowances reset on the 1st of each month, and you can track usage in your dashboard. If you need more without changing plan, there are message token add-ons.

Question 13

What languages does Chat Thing support?

Accepted Answer

Your bot can understand and reply in 50+ languages, and it detects the customer's language automatically — there's nothing to configure and no need for a separate bot per language. You can write your data sources and system prompt in English and your bot will still answer a customer in Japanese, German or Spanish in their own language. Quality is strongest in widely-spoken languages. If you're relying on a specific one, the free plan is enough to test it.

Question 14

Can I test my bot's answers before customers see them?

Accepted Answer

Yes — bot testing lets you build a suite of test cases and run them against your bot, checking each response for factual accuracy, similarity to an example answer, relevance, and any requirements you define. Each run snapshots the bot's configuration so you can compare prompts and models. Bot testing is available on the Enterprise plan. On any plan you can chat with your bot directly from its dashboard to spot-check answers before going live.

Question 15

Which plans include Slack, Discord, WhatsApp and Telegram?

Accepted Answer

Chat platform channels — Slack, Discord, WhatsApp, Telegram and the API channel — are available on the Standard plan and above. The web channel is available on every plan including Free, so you can embed your bot on your website, share its direct link, or drop it in as an iframe without upgrading. See what is a channel for setup guides.

Question 16

Do you integrate with Microsoft Teams?

Accepted Answer

Not currently. Chat Thing supports the web channel, Slack, Discord, WhatsApp, Telegram, email and a general-purpose API channel. If you need Teams, the API channel can be used to build the integration yourself, and we'd like to hear about it — email support@chatthing.ai so we can weigh it up.

Question 17

Can I use Chat Thing for sensitive topics like health information or charity signposting?

Accepted Answer

Yes, and it's a common use case — signposting to services, answering questions about what you offer, and pointing people to the right resource around the clock. A few things we'd recommend for this kind of bot: Set expectations in the system prompt. Tell the bot to share general information and signpost, and to avoid anything resembling diagnosis or personal advice. Set up escalation before you launch. Add the Talk to a Human power-up so anyone who needs a person can reach one. Say it's a bot. People share more when they think they're talking to a professional. A clear note in your welcome message helps. Don't use it for crisis support. A chatbot won't reliably pick up on distress the way a trained person will. Signpost to a helpline instead. You remain the data controller for anything your customers type, so apply your usual safeguarding and data protection policies. Our Privacy Policy and DPA cover how we process it.

Question 18

Does the chat widget match my brand colours?

Accepted Answer

Yes. You can set primary and secondary colours, choose a light or dark theme, and upload your own assistant and user avatars, all from your bot's Web channel settings under "Appearance and branding". There are quick presets to start from, and custom CSS if you want full control. See widget customisation.

Question 19

Can I connect Chat Thing to Claude, ChatGPT or Cursor?

Accepted Answer

Yes — Chat Thing has an MCP server at https://app.chatthing.ai/mcp, so you can manage bots, data sources, power-ups and channels from Claude, ChatGPT, Cursor or any MCP-capable client. In Claude, add it under Settings → Connectors → Add custom connector, leaving the OAuth fields blank. Full walkthrough in getting started with MCP.

Question 20

What file types can I upload?

Accepted Answer

You can upload .pdf, .csv, .json, .md, .html, .mht, .docx, .txt, .zip and .epub files as a data source. Text is extracted and indexed so your bot can answer from it. Images inside documents aren't read as images, though text in a PDF is. Upload a .zip to add many files at once. See file data sources.

Question 21

Can the bot change my plan or account settings for me?

Accepted Answer

No. Our support bot can explain how something works and point you to the right screen, but it can't make changes to your account — including plan changes, billing, or cancellations. Those are always done by you from your billing page, so nothing changes without your explicit action. If a bot ever tells you it has made an account change, treat that as a mistake and check your dashboard or your email for confirmation. Bots you build can take real actions through power-ups — looking up an order, updating a record — but only for the integrations you connect and configure yourself.

Question 22

Can my bot hand off to a real person, and which plans include it?

Accepted Answer

Yes, on the Standard plan and above. Standard and up: full human takeover. A customer asks for a human, the conversation appears in your Conversations dashboard, a team member claims it and chats live, then hands back to the bot. Web channel only (including embedded widgets and iframes). Basic: the Talk to a Human power-up can still collect the customer's email and notify your team, so nothing gets lost — but no one can take over the live conversation. Free: power-ups aren't included, so neither escalation route is available.

Question 23

Is my data used to train AI models?

Accepted Answer

No — and for chat messages it goes further than that. Your conversations aren't just excluded from training, they aren't stored by the AI provider at all. Chat Thing never trains on your data. Your content, your data sources and your customers' conversations are never used to train AI models. This is committed to in our Privacy Policy. Neither do the AI providers. We run on the commercial APIs from OpenAI, Anthropic, Google and others, routed through OpenRouter. We have Zero Data Retention enforced account-wide, which means your prompts and your bot's responses are processed in memory and discarded. Nothing is retained on the provider side — so there's nothing to train on, nothing to leak, and nothing to hand over later. Zero Data Retention is enforced at the account level, so it applies to every bot, on every plan, on every model, without you configuring anything. One exception worth knowing: if your bot uses the web chat's voice features, audio and reply text are sent to OpenAI's API directly rather than through OpenRouter. That content is still never used for training, but OpenAI's standard API retention for abuse monitoring applies, so it isn't zero-retention. Text-only bots are unaffected. Need this in writing for a security review? We can provide a DPA — email support@chatthing.ai.

Question 24

Don't the LLM providers train on data by default unless you opt out?

Accepted Answer

That's true of the consumer products — ChatGPT, the Claude app, Gemini — where your chats can be used to improve the model unless you go and turn it off. It's a fair thing to ask, and it's the most common misunderstanding we hear. It isn't how the API tier works, and the API is what Chat Thing is built on. API terms exclude customer content from training by default: OpenAI's terms of use (section 3.c) state they "do not use Content that you provide to or receive from our API to develop or improve our Services", and Anthropic and Google publish equivalent commitments. On top of that default, we enforce Zero Data Retention on our own account for chat messages. Requests are only ever routed to model endpoints that don't retain prompts — and if a compliant endpoint isn't available, the request fails rather than quietly falling back to one that would store your data. The one path that doesn't go through OpenRouter is the web chat's voice features, which call OpenAI directly. Training is still excluded there; standard API retention applies.

Question 25

Do I need to change any setting to keep my data private?

Accepted Answer

No. There's nothing to opt out of and nothing to switch on. Every protection is enforced at the platform level and applies to all bots, all plans and all models from the moment you sign up — including the free plan. You can see every third party that processes your data, what each one does with it, and how retention differs between chat and voice, on our sub-processors page.