Replit chatbot for support: what it takes to build

Replit doesn’t come with a support chatbot for your users, but its Agent will build you one. Replit’s chatbot builder page has you describe what the chatbot should do and connect OpenAI or another model provider by pasting your API key. On the Core plan and above, Replit AI Integrations lets the chatbot call models with no key of your own, billed to your Replit credits. We read Replit’s docs, guides and pricing on 23 September 2026.

Deacon is an AI customer support agent you add to a Replit app with one script tag, so there’s nothing to build or host. It answers your customers in seconds from your own docs, says so when it doesn’t know, and shows you what they ask about most. The free plan gives you 50 answers a month, needs no card and has no end date.

What building a support chatbot on Replit takes

Replit’s own Website Q&A RAG Chatbot guide shows what sits under a chatbot that answers from your site. It takes an OpenAI API key saved in Replit’s Secrets, a PostgreSQL database with pgvector to hold the embeddings, LangChain to split your pages and search them, and a script that reads your sitemap.xml and fetches every page it lists. Then comes a frontend and a deployment. Agent can write these parts for you, but they’re the parts either way.

The guide ends once the chatbot answers and is deployed. A chatbot that handles support needs four more things, and you ask Agent for each one and maintain it after.

1. A reply for when the search finds nothing, so the model doesn’t fill the gap with a guess.
2. A way for a user to reach you, and for you to hear about it, such as an email when someone asks for a person.
3. A log of what people asked, with the questions it couldn’t answer picked out, so you know what to add.
4. A fresh run of the scrape whenever your docs change, so the answers keep up.

What a Replit chatbot costs to run

The model is the first cost. With your own key, the provider bills you directly. With Replit AI Integrations, Replit bills the public API price to your Replit credits, and it needs the Core, Pro or Enterprise plan. On the free Starter plan it’s switched off, so you bring a key.

The plan is the second. Core costs $20 a month, or $18 a month billed yearly. Replit’s Core page says its monthly credits pay for Agent, publishing, network transfer, database storage and AI Integrations, so the chatbot’s model calls come out of the same credits Agent builds with. The free Starter plan gives you one published app, which goes down after 30 days and carries a “Made with Replit” badge until you move to Core.

The database is the third. Replit bills a production database by compute time and storage, and its August 2026 price list puts those at $0.16 per compute hour and $0.35 per GiB a month. The development database you build with is free.

On Core and Pro, usage past your monthly credits is billed as you go. You can cap that, but Replit’s spending guide says that when you reach a usage limit, usage-based services are blocked until the next billing cycle or until you raise the limit. So a capped chatbot goes quiet in a busy month, and an uncapped one keeps billing.

What Deacon does on a Replit app

You chose Replit to ship without running infrastructure. A support chatbot of your own brings a database, a scraper and a monthly bill back. Deacon answers your users in seconds, at any hour, from the docs and pages you choose and in whatever language each person writes in. It runs on Deacon’s servers, so there’s no key to manage and nothing to host, and its answers use none of your Replit credits.

When your docs don’t cover a question, Deacon says it doesn’t know rather than making something up, and asks for the user’s email. If they leave it, or ask for a person, you get an email. Reply from Deacon’s dashboard, and if they still have the chat open, your reply appears there under your name.

Then Deacon gets better. Every question your docs didn’t cover waits in the dashboard with an Answer this button, and what you write becomes part of what Deacon knows. It can answer the next person who asks, so the more you answer, the less often you’re needed. When one of your pages changes, you fetch it again and retrain from the dashboard, with no scraper of your own to run.

You also hear from everyone who asks, not only the few who email. Every conversation is saved and searchable by what users asked. Topics groups the product questions, counts how many people asked each one and labels most of them, such as a feature ask, a how-to or a bug report. That’s customer feedback you never had to ask for, and it shows you what to fix and what to build next.

Every feature is on every plan, the free one included, and there’s no badge to pay to remove. Nothing is billed past your plan’s answers, no plan upgrades itself, and every plan answers with the same model, so no credits run out faster on a better one. The pricing page has the paid plans.

Adding Deacon to a Replit app

It takes a few minutes, and Deacon and Replit has every step.

1. Sign up, then add your docs or your site in Knowledge Sources.
2. On the Install page, list your app’s replit.app address, and your own domain if you’ve connected one. The preview address is temporary, so test on the published one.
3. Copy your script tag from the Install page, paste it into the Agent chat and ask Agent to put it just before the closing body tag of the page your app serves, or in its layout, so it loads on every page.
4. Republish, because changes in the Project Editor don’t reach your live app until you do. Deacon’s Install page shows when it has seen your script on a live address, with a Check again button.

If the chat never appears, look for a content security policy. Replit’s security checklist suggests one that lets no outside script run, and the Deacon and Replit page lists what to add to it.

Questions about Replit chatbots

Does Replit have a chatbot builder?

Replit’s Agent is the builder. Its chatbot builder page has you describe the chatbot, then Agent writes the files, connects the model and deploys it. What it builds is your code, so answering only from your docs, handing over to you and logging what people asked are parts you ask for and maintain. Lovable chatbots and Bubble chatbots cover building one on those platforms.

Can I build a chatbot on Replit’s free plan?

Yes, with your own API key, since Replit AI Integrations is switched off on the Starter plan. Starter publishes one app, which goes down after 30 days and can be republished, and it shows a “Made with Replit” badge. Core, at $20 a month or $18 billed yearly, lifts both limits.

Do I need an OpenAI key for a Replit chatbot?

Not on Core, where Replit AI Integrations calls OpenAI and other providers with Replit’s own credentials and bills the public API price to your credits. On Pro and Enterprise an admin switches it on first. On Starter, or if you’d rather be billed by the provider, you save your own key in Replit’s Secrets.

Can I keep a chatbot I built on Replit and add Deacon?

Yes. Deacon is a script tag and changes nothing in your app’s code or database, so what you built keeps working. Keep your own chatbot for things only your app can do, such as working with a signed-in user’s records, and let Deacon answer everyone’s questions from your docs. Deacon never looks into a user’s own account or data, and declines those questions.

Does Deacon use my Replit credits?

Only for the one request that asks Agent to place the tag. After that, Deacon’s answers run on Deacon’s servers and come from your Deacon plan, so your Replit credits go on your app.

What happens when Deacon’s free answers run out?

The chat keeps working. A user who asks gets a short reply saying Deacon can’t look into that one right now, and is asked for an email unless they’ve already left one in that conversation. You can still read and reply to every conversation, nothing is billed and your plan doesn’t upgrade itself.