Add customer support to the app you built in Cursor — one script tag
You wrote and shipped a real product from your Cursor codebase. Real products get real "how does this work?" questions. HelpShelf answers supported questions from your docs and README, installs with one snippet, and stops clearly when it cannot verify an answer. With tested Telegram, the visitor can choose to send a handoff. (This adds support to your app; it is not Cursor's support desk.)
30-second demo
support agent live in one command
Run the command, let HelpShelf scan and publish your support surface, then connect the printed MCP URL in Cursor.
npx helpshelf --source=cursorCan you add customer support to a Cursor app?
Yes. An app built in Cursor is a normal web app deployed on your own host. Point HelpShelf at your live URL or docs, optionally connect GitHub for README content, then paste one script tag into the HTML head or root layout. The widget gives supported answers with citations; when it cannot verify an answer, it can offer a visitor-confirmed handoff only if your tested Telegram destination is reachable. There is a Free plan, so you can ship it before you have a single paying customer.
Cursor helped you write the app. It did not write your support replies.
You used Cursor to move fast — autocomplete, refactors, whole features shipped in an afternoon. But once the app is live, the questions arrive in a channel Cursor never touches: your inbox. And answering "where do I reset my password?" for the hundredth time is not why you opened the editor.
- The same onboarding, billing, and 'how do I' questions land in your inbox daily, pulling you out of the editor and back into manual replies.
- Your product knowledge is scattered across a README, in-repo docs, and your own head — users have nowhere to self-serve, so every question routes to you.
- Bolting on a traditional help desk means a new dashboard, macros, and seat fees — overhead that dwarfs the app you just shipped from Cursor.
- Generic AI chatbots invent endpoints and features your app does not have, and a confidently wrong answer is worse than no answer at all.
Answers that sell
You shipped it. Now it has to sell.
Most questions on a new Cursor app are not support tickets — they are buying questions from visitors deciding whether to pay. HelpShelf answers them on the spot with cited answers drawn from your own content, nudges the hesitating visitor with a proactive teaser, and can offer follow-up after a confident answer. A lead is recorded only after the visitor submits a valid email. And because every HelpShelf site ships llms.txt and a live MCP server, the AI assistants your buyers use can inspect your product sources when those surfaces are linked or configured. Support questions? The same agent handles those too, included.
Sales mode where it counts
Auto-pinned to your pricing and marketing pages at onboarding — the agent sells there and supports everywhere else.
A nudge for the hesitating visitor
A proactive teaser bubble invites the question that is stopping them from signing up.
Leads, not closed tabs
After a confident answer the agent offers a follow-up — the visitor’s email lands in your dashboard and Telegram.
Sales Pulse on your dashboard
Sales conversations, captured leads, top buying questions, and AI-shopper traffic in one view.
From Cursor build to live support in three steps
No help desk migration. No new platform to learn.
Point it at your shipped app
Paste your live app URL or docs and HelpShelf indexes them into a knowledge base — no source code needed. Optionally connect the GitHub repo behind your Cursor project so the agent also learns from your README and in-repo docs.
Paste one snippet into your codebase
Copy the HelpShelf <script> tag and add it to your app's HTML — index.html for a static build, or your root layout / _document for a framework app. Commit, push, and your usual deploy ships the widget on every page.
Ship it and let the agent answer
Users get supported answers with citations from your own content. If the agent cannot verify an answer and tested Telegram is reachable, it can offer a human handoff; when the visitor accepts, Telegram receives bounded context. Every conversation remains reviewable in your HelpShelf inbox.
Built for how you shipped it
Fast to install. Grounded in your content. Stops safely when it cannot verify an answer.
One script tag to install
Paste your URL, drop the snippet into your head tag, deploy. No help-desk setup, no macros, no new backend to maintain — it fits the way you already ship from Cursor.
Answers from your repo and docs
The agent learns from your live app, your docs, and — if you connect GitHub — your README. Replies are grounded in what you actually built, with citations users can click.
Lives in your codebase
It is one script tag in your own HTML. No proprietary SDK, no framework lock-in — works in any app you deploy from a Cursor project, from a static site to Next.js.
Cites or stops
When an answer cannot be verified, the agent stops and can offer a visitor-confirmed handoff if your tested Telegram connection is reachable.
Free while you find traction
Start on the Free plan with a live widget, AI answers, and your own help center. Pro starts at $49/mo with 10,000 AI searches and 2,000 AI chats each month when volume picks up.
Stays current as you ship
Refactor a feature in Cursor and ship it, then re-sync your content in HelpShelf and the agent relearns. Your docs and your answers stop drifting apart.
Perfect for Cursor builders
Whatever you shipped, your users will have questions
Indie devs shipping fast
You build features in Cursor faster than you can write docs. Let the agent handle supported repetitive questions so you stay in the editor, not the inbox.
Solo SaaS founders
First users can ask supported questions and inspect cited answers; when the agent cannot verify one, the visitor can confirm a handoff to your tested Telegram destination.
Client and freelance builds
Hand off a Cursor project with a working AI support agent already wired in. One snippet in the head, and the delivery looks far bigger than the effort.
Internal tools and dashboards
Drop a self-serve help agent into the internal app your team uses so supported 'how do I…' questions have a cited answer path before a DM.
Widget installation
To get started
Supported answers
Frequently asked questions
Yes. An app you built in Cursor is a standard web app you deploy yourself. Point HelpShelf at your live URL or docs, then paste one script tag into the app's HTML head or root layout. The widget gives supported answers from your content, and there is a Free plan to start. (This adds support to the app you made, not to Cursor the editor.)
Create a free HelpShelf account, add your live URL or docs as a source, customize the widget, and copy the embed snippet. In your Cursor codebase, paste that <script> tag into your app's HTML — index.html for a static build, or your root layout / _document for a framework like Next.js — then commit and deploy. The widget is live for every visitor on your next deploy.
No. HelpShelf answers from your content — your live app, help docs, or pasted text — not your source code. You can optionally connect the GitHub repo behind your Cursor project so the agent also learns from your README and in-repo docs, but that is a bonus, not a requirement.
When the agent cannot verify an answer, it says so rather than presenting an invented feature, endpoint, or price. Every confident answer is grounded in your sources with citations. If your tested Telegram connection is reachable, the visitor can review and confirm a human handoff.
Yes. The Free plan includes a live widget, AI answers, and your own help center — enough for most side projects and early MVPs. Pro starts at $49/mo with 10,000 AI searches and 2,000 AI chats each month when you start handling real volume.
When you change features or copy and ship from Cursor, re-sync and review the relevant app content, docs, or README in HelpShelf. New answers use the updated published sources; conversation records and knowledge gaps show what needs attention.
Build with a different tool?
HelpShelf adds an AI support agent to any app, whatever you built it with.