How do I add an AI chatbot to my site without managing my own API keys?
Short answer: If you have a developer or an AI coding tool building the feature for you, Netlify's AI Gateway lets your site's code call AI models (OpenAI's, Google's, Mistral's and others) directly, without you creating a separate account with each AI provider or managing your own API keys. It's billed through the same account credits as everything else on Netlify, priced per model rather than a flat fee. If you don't have a developer involved at all, an embedded chatbot vendor like Intercom or Tidio, which you can set up yourself with no code, is usually the simpler and faster starting point, and this page is honest about which situation you're actually in before you pick a path.
You want an AI feature on your site, a chatbot or a text summary tool, and you don't want to be the one wrangling API keys and provider accounts to get it there.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
This page may contain affiliate links from which Blur Studio earns commissions.
Adding AI to a website usually means picking an AI provider, creating an account, generating an API key, and wiring that key securely into your code so it isn't exposed to visitors, which is more setup than it sounds for anyone without a developer background. Netlify's AI Gateway removes the provider-account part of that: your site's code can call supported AI models directly through Netlify, with Netlify managing the underlying authentication, so nobody needs to create a separate OpenAI or Anthropic account or handle a raw API key at all.
Who this page is actually for
Be honest with yourself about this before reading further: AI Gateway is a tool for a developer or an AI coding agent to build on, not a chatbot you install yourself with no code. If you want a chatbot widget live on your site this week with zero development work, an embedded vendor like Intercom or Tidio, both of which you set up directly through a dashboard, is the right starting point, not this. AI Gateway is the right tool when you (or whoever's building it for you) want a custom AI feature, not a generic chatbot widget, something that answers questions using your specific business knowledge, summarizes a document a customer uploads, or does something a stock chatbot vendor doesn't offer out of the box.
How it actually works
Once your site is built on Netlify, its code (written by a developer, or generated through an AI coding tool like Claude or ChatGPT connected via Netlify's MCP Server) can call AI models through the Gateway without embedding a provider-specific API key anywhere in that code. Netlify's changelog shows it actively keeps adding new models: OpenAI's GPT-6.1 Terra landed October 26, 2026, Google's Gemini 4 Argon on October 13, and Mistral Large 4, with a large 524K-token context window, the same day, all confirmed directly on Netlify's own published changelog. Usage is priced per model within your existing Netlify credit pool (Free at $0 with 300 credits, Personal at $9/month with 1,000, Pro at $20/month with 3,000, Enterprise custom), rather than a flat monthly chatbot fee.
What this actually costs in practice
Because usage is metered per model call rather than a flat subscription, the real cost depends entirely on how much the feature gets used and which model it calls. A simple FAQ-answering feature that gets used a handful of times a day costs a small, often barely noticeable slice of your monthly credit pool. A chatbot that gets heavy daily traffic, or one built on a larger, more expensive model for every response, can add up meaningfully faster. The practical move before committing to a specific model: have whoever's building the feature test it on a smaller, cheaper model first and only upgrade to a larger one if the responses genuinely need it, since the quality difference isn't always worth the cost difference for a straightforward Q&A feature.
Usage can also burn through credits in a way that pauses your whole site if you're not watching it, since Netlify's credits are shared across hosting and AI usage rather than siloed separately. That's worth flagging specifically to whoever builds this for you: ask them to set a sensible usage cap or rate limit on the AI feature so a spike in traffic, or a bug that calls the model in a loop, can't quietly take down your entire site's hosting along with it.
The alternatives, and when they actually win
Embedded chatbot vendors like Intercom or Tidio win when you want something live fast with no development work: you add a script tag, configure some canned responses or connect a knowledge base through their own dashboard, and you're done in an afternoon. They're built for breadth (reasonable AI-assisted answers across many kinds of business) rather than depth on your specific situation. Calling the OpenAI API directly, without Netlify's Gateway in between, wins if you already have a developer comfortable managing API keys and provider billing separately, and you want direct control without any layer in between. AI Gateway sits in the middle: less setup than managing your own provider account, but still something a developer or AI coding tool needs to actually build, not a self-serve widget.
What I'd actually tell you
If what you actually want is "customers can ask my site a question and get a sensible answer," start with an embedded vendor and see if it's genuinely not good enough before building anything custom. Most small businesses overestimate how much their chatbot needs to be custom versus how much it needs to just exist and work reasonably well. Where a custom build through something like AI Gateway earns its complexity is when the feature needs to know things a generic vendor can't easily be taught, your specific pricing logic, your actual inventory, a workflow unique to how your business runs, in which case the flexibility is worth the extra development step.
When to skip this entirely
Skip building a custom AI feature if you don't have a developer or a reliable AI coding workflow already in place, since the setup, even with Netlify handling the provider-account part, still needs someone to write and maintain the actual code, and usage can burn credits unexpectedly if nobody is watching it. It also needs a developer to build the feature in the first place; this isn't a configuration toggle, it's a real piece of software that someone has to design, test and maintain.
A practical note on testing before launch
Before putting a custom AI feature in front of real customers, run it through a handful of deliberately odd or off-topic questions, not just the straightforward ones you expect people to ask. AI features that work perfectly on the three test questions used during development can respond strangely, or confidently wrong, to something unexpected a real visitor types in. A short list of edge cases tried before launch, including questions genuinely unrelated to your business, catches most of the embarrassing failures before a customer does.
How I'd approach this
When someone comes to me wanting "AI on the website," my first question is always what specific question or task they want it to handle, because that answer usually reveals whether a ready-made chatbot vendor already covers it or whether it genuinely needs something custom. If it's the latter, building on something like Netlify's AI Gateway is a sensible foundation, and I'd set it up with a usage cap from day one so a traffic spike or a bug never threatens the rest of the site's hosting.
Questions people ask
Do I still need an OpenAI or Anthropic account?
No, that's the specific thing AI Gateway removes. Your site's code calls supported AI models through Netlify directly, without a separate provider account or a raw API key embedded in your code.
Can I set this up myself with no coding?
No. This is a tool for a developer or an AI coding agent to build on, not a self-serve chatbot widget. If you want something live today with zero development, an embedded vendor like Intercom or Tidio is the right starting point instead.
How is this billed?
Through the same account credit pool as the rest of your Netlify usage (300 credits free, 1,000 on Personal at $9/month, 3,000 on Pro at $20/month), priced per AI model call rather than a flat chatbot subscription fee.
Can an AI feature accidentally take down my whole site?
If usage isn't capped, yes, since credits are shared across hosting and AI usage. Ask whoever builds the feature to set a sensible rate limit or usage cap so a traffic spike or a bug calling the model repeatedly can't drain the credits your site needs to stay up.
Which AI models are supported?
Netlify actively adds new ones; its own changelog shows OpenAI's GPT-6.1 Terra, Google's Gemini 4 Argon and Mistral Large 4 all added in October 2026 alone, each usable without you managing a separate account for that provider.
When should I just use a chatbot vendor instead of building this?
When what you actually need is generic question-answering or canned responses rather than something tied to your specific business logic, pricing or inventory. A vendor like Intercom or Tidio covers that case faster and without development work.
Is a cheaper AI model noticeably worse for a simple chatbot?
Often not for straightforward Q&A. Testing on a smaller, cheaper model first and only upgrading if the responses genuinely fall short is usually the more cost-effective starting point than defaulting to the most expensive model available.