Use your own AI, on any plan.
SupportVia is a website chatbot you can run with your own OpenAI API key, or your own Claude, Gemini or other OpenAI-compatible account, on every plan and during the free trial. You keep the chat widget, the shared inbox, the “Read more” links and the handover to your team, and the same rules apply: answers only from your pages, and a person when it isn’t sure. It can answer up to 1,000 conversations a month on Starter, up to 5,000 on Growth and no monthly cap on Pro, and its answers never use your plan’s AI conversations.
Free for 7 days or 50 AI conversations. No card needed. Live in about ten minutes.
- Customer: How long does a kitchen design take?
- Oak and Pine Studio: Most kitchen designs take 3 to 4 weeks from the first visit to final drawings. (Read more: Kitchen design)
- Customer: Can you work with my builder’s plans from 2009?
- Oak and Pine Studio: I’m not sure about that. Want me to get someone from the team?
Providers you can connect
Each workspace can connect one AI account from any of these:
- OpenAI
- Claude
- Google Gemini
- OpenRouter
- Groq
- Mistral
- DeepSeek
- xAI
- Together AI
- Hermes Agent (self-hosted)
- OpenClaw (self-hosted)
- Any other OpenAI-compatible API
How connecting works
- Choose your provider and paste your API key.
- Pick a model from the provider’s live list of models.
- SupportVia tests the connection before it saves it, so a wrong key or model shows at once.
Your key stays on our server. It’s never sent back to the browser: the dashboard only shows that a key is set.
The same rules around your model
Your model writes the replies. SupportVia decides what it reads and what happens next, so the chat behaves the same whichever model you pick:
Only your pages
Only the knowledge that matches each question is sent to the model, with the rule that prices, policies and contact details come from there.
Read more links
Answers link to up to five pages they came from.
Says when it isn’t sure
When nothing matches, the model is told so, and asked to offer your team instead of guessing.
People win
Once a teammate replies, the AI stops. If your provider errors or times out, the chat goes to your team with an apology to the customer.
Smaller models follow these rules less reliably than larger ones. Our guide to chatbots that give wrong answers has questions to test any model with.
How much your own AI can answer
Your own AI works on every plan, and during the free trial so you can try it: up to 1,000 conversations a month on Starter, up to 5,000 on Growth and no monthly cap on Pro. A conversation counts once a month, however many replies it takes.
Its answers never use your plan’s AI conversations, which are for SupportVia’s built-in AI, and chats your team answers are never counted. Hourly fair-use limits apply to every answer. You pay your provider for what your model uses, at their rates.
| Plan | A month | AI conversations | Own AI conversations | Websites | Teammates |
|---|---|---|---|---|---|
| Starter | $29 | 150 | 1,000 | 2 | 2 |
| Growth | $79 | 500 | 5,000 | 5 | 5 |
| Pro | $199 | 1,500 | No monthly cap | 15 | No limit |
Prices are in US dollars or rupees (₹), plus taxes where they apply. When the month’s cap is used up, new chats go to your team until the 1st (UTC). You set how many new chats one visitor can start with your own AI each day (10 by default, or off), so one visitor or a bot can’t use up the month. With no trial or plan, the connection is kept but paused, and your team answers. For how other tools charge for AI, see the AI chatbot pricing study.
Self-hosted agents: Hermes and OpenClaw
If you run Hermes Agent or OpenClaw yourself, the dashboard takes you through the setup step by step, with a prompt to paste into your agent so it follows the same rules.
Your own agent
Built your own agent? Give it an OpenAI-compatible endpoint at a public https address, and connect it to a workspace as “Other”. It gets the conversation and the matching knowledge, and SupportVia handles the widget, the inbox and the handover.
A fully custom HTTP agent (JSON in, a reply, a handoff and sources out) is also possible, but it’s a setting for the whole server, not for one workspace.
Questions about a chatbot with your own OpenAI or Claude key
Can I use my own AI on Starter?
Yes. Your own AI works on every plan and during the free trial: up to 1,000 conversations a month on Starter, up to 5,000 on Growth and no monthly cap on Pro. Its answers never use Starter’s AI conversations.
Which model should I pick?
Every model gets the same rules, but smaller models follow them less reliably. Try your customers’ common questions in the playground, which uses the AI you’ve connected and stores nothing, and keep the model whose answers and handovers you trust.
Is my key shared with anyone?
No. It’s stored on our server, used only for your workspace (its chats, the playground and the text the AI writes for your chat), sent only to the provider it belongs to, and never sent back to the browser.
Can each workspace use a different provider?
Yes. Each workspace connects its own AI account, on any plan, so one client can use OpenAI and another Claude.
Can I plug in a fully custom agent?
Per workspace, yes, as an OpenAI-compatible endpoint at a public https address. A raw HTTP agent with its own format is a setting for the whole server, not for one workspace.
More from SupportVia
Answer every customer. Start free today.
Add your website, check what it learned, and paste one line. Your team sees the first conversation in the inbox.