SaaS support

AI customer support for SaaS,
grounded in your docs

For founders, product engineers, and small customer-success teams who already wrote the answers in a help center, changelog, or product guide. Heeya retrieves those passages, generates a reply, and ships as a website widget so you do not have to run your own retrieval pipeline.

Create a free Heeya account Try Heeya's live assistant

Built for documentation-heavy B2B SaaS and AI products. Current plans are on Heeya pricing.

Product docs assistant

Illustrative conversation

How do I invite a teammate as an admin?

Open Settings, then Members, then Invite. Choose Admin before you send the invite. Admins can manage billing and sources. That is from your Team permissions guide.

Can they also change the workspace plan?

Yes, according to that same guide: Admin is the role that can change the plan. If you want someone to invite people without billing access, use Member instead.

What AI customer support for SaaS actually is

In this context, AI customer support for SaaS means a chat agent on your product or marketing site that answers from your documentation: onboarding guides, feature pages, setup troubleshooting that you already published, plan and policy copy, and the questions that arrive after hours when nobody is in the queue.

Heeya's place in that job is narrow on purpose. You add sources, write guidance, test in the dashboard, then paste one script onto the site. Visitors talk to the widget. You can later read the conversations and helpful/not-helpful marks. That is the product: a retrieval-backed agent plus an embeddable chat UI, not a helpdesk and not a custom model you train.

If you need the longer onboarding-and-support narrative, including activation and CS workflow, that lives on the AI chatbot for SaaS onboarding and support guide. This page is the commercial fit check.

The support jobs a SaaS team already has

These are the questions a docs-grounded agent can attempt. They are only as good as the source you loaded for each job.

Onboarding questions

First-session setup, where a setting lives, which step comes next. Useful when the answer is already in a getting-started guide, not when the user needs a human to inspect their account.

Feature documentation

"Does this exist?", "Where is it?", "What does this toggle do?" The agent can retrieve the matching article. It cannot demo the UI inside your app.

Setup troubleshooting covered by docs

Known error messages, required fields, and configuration order that you already wrote down. Novel bugs and account-specific failures still need a person.

Plan and policy explanations

What a plan includes, how billing periods work, refund windows, usage limits: only if that copy is in the knowledge base. The agent will not change the plan for the visitor.

After-hours questions

The widget is available whenever the page is loaded. That is coverage of documented questions, not a support SLA and not a promise that every night-time question has an answer in your docs.

Hand-off when docs run out

If you enable the form tool, the agent can show a contact form in the chat so the visitor leaves details in one block. That is capture, not automatic routing to a named owner.

How it works

Five steps that match the product as it exists today. Nothing here assumes a native app integration.

1. Sources

Add documents, pages, or Q&A

Upload PDF, DOCX, PPTX, or TXT. Paste a site URL, discover pages from the sitemap or llms.txt, and scrape the ones you select. You can also add question-and-answer pairs, including from a spreadsheet.

2. Guidance

Configure tone and branding

Write system guidance. Set the agent name, welcome and popup copy, up to three suggested questions, bubble and chat colors, an optional logo, and English or French as the deployment language.

3. Test

Ask the questions your users ask

Use the dashboard chat against the same retrieval path before anything goes on the site. If a reply is thin, the fix is usually a missing or stale source, not a new integration.

4. Embed

Paste the script on the site

Copy the snippet from the connect page and place it on the HTML page where visitors should see the widget. There is also a full-page assistant URL if you want a standalone chat link.

5. Review

Read conversations and marks

Analytics lists conversations so you can open the transcript. Visitors can mark a reply helpful or not helpful. Use that to see which articles are missing, not as an automated quality score.

The integration is a script tag, not a native package

Heeya is a drop-in web widget. After you create an agent, the connect screen gives you a snippet in this shape:

<script async src="https://heeya.fr/agent/YOUR_AGENT_ID/embed.js" data-agent-name="Support"></script>

Put it on any page that can load a third-party script, typically before the closing body tag. In a React or Next.js app that still means loading this script, not installing a Heeya package. Colors and the logo are configured on the agent; they are not something you pass as an npm prop.

If you want a page that is only the chat, the same agent also has a public assistant URL: https://heeya.fr/agent/YOUR_AGENT_ID/assistant.

That is the verified surface. There is no public Heeya HTTP interface for your backend, no native Slack or WhatsApp channel, and no HubSpot connector to claim here.

What docs-grounded answers improve, and what they do not

Heeya uses retrieval-augmented generation. If there is chat history, the follow-up is rewritten into a standalone question, then similar chunks are retrieved from that agent's collection and injected as context. The model is told to use that context and to say when the information is not there. File names and page URLs are available to the model in that context. Visitors do not get a separate citation panel in the widget.

What this improves

Answers can track your current docs instead of a generic model's memory of other products.
Updating a source and re-ingesting it is how you change an answer, rather than rewriting bot flows by hand.
When the chunk is missing, a well-behaved reply is "I do not have that," plus, if you enabled it, a form.
For the architecture behind this pattern, see RAG expertise and RAG for customer service.

What this does not promise

Retrieval reduces unsupported answers. It does not make them impossible.
Stale, contradictory, or incomplete docs produce stale, contradictory, or incomplete replies.
The agent does not execute refunds, role changes, or other account actions.
There is no certified "zero error" mode, and no SOC 2 badge to display on this page.

A practical knowledge-source checklist for SaaS

Load the documents people already quote in tickets. Skip secrets. For how to structure a corpus, use the knowledge base engineering guide.

Load first

Help center or docs site (scrape selected URLs)
Getting-started and setup guides (PDF or DOCX)
Feature reference that matches the current product
Plan comparison, billing, and refund policy pages
Changelog or "what shipped" notes if users ask "do we have X yet?"

Add when the first pass is thin

Top recurring tickets rewritten as Q&A pairs
Integration setup articles you already published
Known error messages and the documented fix
Role and permission matrices
Public status or incident policy, if you want that in chat

Do not upload

Customer PII, ticket dumps with emails, or invoices
Private keys, tokens, or internal admin runbooks
Draft docs that contradict the public product
Retired versions that still rank in retrieval
Anything you would not want a visitor to hear quoted back

Fit, and not a fit

Heeya is a fit when

Most incoming questions already have a written answer in docs or a help center.
You want that answer on the website without standing up Qdrant, embeddings, chunking, and a chat UI yourself.
A small CS or product team will maintain sources and skim conversations.
A contact form in the chat is enough when the agent should stop guessing.

Look elsewhere when

You need the agent to take transactional actions in your product (provisioning, refunds, role changes).
You need native Slack, WhatsApp, HubSpot, or a webhook-based ticket router. Those are not available in Heeya today.
You need a full helpdesk replacement with assignment, collision detection, and human takeover of a live widget session.
You need a public programming interface or a first-party React/Next component library. The embed is the integration.
You need calendar booking or appointment holds. Heeya does not connect to an agenda.

Build a retrieval stack, or use Heeya

Building in-house means owning ingestion, chunking, embeddings, a vector store, prompt policy, evaluation, and a chat front end, then keeping that current as docs move. Using Heeya means you operate the sources, the guidance, and the snippet, and you accept the product's current limits: widget-only delivery, form-plus-search tools, no ticket router.

The cost, time, and criteria for that decision are already written. Use the custom AI chatbot build vs buy guide instead of a second matrix on this page.

Pricing

Heeya is sold as a subscription with a free starting tier. Plan names, message limits, and amounts change on the pricing page, so they are not copied here. Open current Heeya plans for the live grid. For market-wide cost ranges, see how much an AI chatbot costs. To model your own ticket volume, use the AI chatbot ROI calculator.

Create a free Heeya account

FAQ about AI customer support for SaaS

Does Heeya replace Zendesk, Intercom, or a full helpdesk?

No. Heeya is a docs-grounded support agent for your website. It answers from the sources you add and can show a contact form in the conversation. It does not replace a ticketing product, assign conversations to humans, or take over an existing helpdesk.

Will the agent invent answers that are not in our documentation?

Retrieval-augmented generation reduces that risk by injecting retrieved passages into the prompt and instructing the model to say when the information is missing. It does not remove the risk entirely. You should still test answers against your docs and keep sources current.

Can I install a React or Next.js package?

No. Heeya ships as a drop-in website widget: one script tag that loads the chat UI. You can place that script on any HTML page, including a React or Next.js site, the same way you load other third-party tags. There is no public Heeya package to install.

Can the agent change a plan, issue a refund, or run actions in our product?

No. The agent searches your knowledge base and, if you enable it, displays a form so a visitor can leave details. It does not call your product, update accounts, or route tickets.

What sources can a SaaS team add?

You can upload PDF, DOCX, PPTX, and TXT files, scrape selected pages from your website, and add question-and-answer pairs. Website discovery can use your sitemap or an llms.txt index when those exist.

How does Heeya compare with Chatbase?

They are both document-grounded chatbot builders. The dedicated comparison lives on the Heeya vs Chatbase page. This page covers whether a docs-grounded widget is the right shape of product for a SaaS support team.

Where are the current plans and limits?

Current plans are listed on the pricing page. This page does not list amounts, because those can change. You can create a free Heeya account from the registration page.

What language does the agent use?

You set the agent's deployment language to English or French. That language drives the widget UI, default greeting, and form labels. The model is also instructed to answer in the language of the visitor's latest message.

Put a docs-grounded agent on the SaaS site you already have

Create an account, add the sources your team already maintains, test the answers, then paste the embed script. That is the path Heeya supports today.