Lovable
Gets the inbox UI and Supabase realtime tables working quickly.
Support · updated 2026-11-03
Can you vibe code an Intercom clone?
Hard: Only a slice A basic chat widget for your own site is a weekend project; a reliable multi-channel inbox with AI answers and mobile notifications is not.
8 h–3 days
MVP build time
$0–$60/mo
Your running cost
$19/mo/seat
Intercom Essential
up to $228
Saved per year
The honest verdict
A chat bubble that sends messages to a table and a small inbox to reply from is very buildable with realtime databases. Intercom's value is everything around it: delivering every message even when the agent is offline, email fallback, multi-channel inboxes, help center search, AI answers grounded in your docs and team workflows. Build a light widget for your own product if you want, but know you are on call for it.
Intercom is customer messaging platform with a live chat widget, shared inbox, help center and an AI support agent. This page is about building your own version for personal or internal use, or to learn how this kind of product works, not about copying Intercom to compete with it. How we grade.
What's easy and what's hard
Easy with AI
- A chat widget that posts messages to a database
- An admin inbox listing conversations
- Realtime updates with Supabase Realtime or a hosted pub/sub
- Basic help articles rendered from Markdown
- Identifying logged-in users in the widget
Hard (where AI won't save you)
- Never losing a message: offline agents, email fallback and reply-by-email threading
- AI answers that are grounded in your docs and know when to hand off
- Embedding a widget safely on other sites without breaking their CSS or security
- Spam and abuse on a public, anonymous chat endpoint
- Mobile push notifications and multi-agent assignment
Intercom feature checklist
Difficulty is for a solo builder using an AI coding tool. Skip means not worth building yourself.
| Feature | Difficulty | Why |
|---|---|---|
| Embeddable chat widget | Medium | Ship it in an iframe or shadow DOM so host page styles do not leak in. |
| Realtime messaging | Easy | Supabase Realtime, Pusher or Ably handle the websocket layer. |
| Agent inbox | Easy | Conversation list, assignment and status open or closed. |
| Email fallback and reply by email | Hard | Needs outbound email plus inbound parsing and threading by conversation ID. |
| Visitor identity verification | Medium | HMAC of the user ID signed by your server so visitors cannot impersonate customers. |
| AI answers from your docs | Medium | Retrieval over your help articles plus an LLM, with clear handoff to a human. |
| Help center | Easy | Markdown articles with search. |
| File attachments | Medium | Signed uploads, type and size limits, malware considerations. |
| Mobile apps and push | Skip | Use email or Slack notifications for a small team. |
Build time and running costs
Your own version vs Intercom, at personal or small-team scale.
Build time
- Working MVP for yourself
- 8 h–3 days
- Polished, relied on daily
- 10–25 days
- MVP time at $50/h equals
- 37 months of Intercom
Intercom price (checked 2026-11-03)
Essential: $19/mo/seat billed annually
Essential is $29 per seat per month, shown as $19 with the current annual or new-customer saving; the Fin AI agent adds $0.99 per outcome. Advanced is $85 per seat.
Source: intercom.com/pricing. Prices change; check before deciding.
| Running cost | Per month | Notes |
|---|---|---|
| Hosting and database with realtime | $0–$25 | Supabase free or Pro includes Realtime. |
| Email in and out | $0–$15 | Email API free tiers plus inbound parsing for replies. |
| LLM API for AI answers | $0–$20 | Optional; depends on conversation volume and model. |
| Total | $0–$60/mo | Saves up to $228 a year vs Intercom Essential per seat; at the top of the range it costs $492 a year more. |
Recommended vibe coding tools
Claude Code
Better for the embeddable widget bundle, identity verification and the email threading logic.
Cursor
Good for integrating the widget into your existing app and iterating on edge cases.
Starter prompt
Paste it into your tool, then ask for a plan before any code. Edit the specifics to match what you need.
prompt.txt
Build a lightweight live chat for my own SaaS (one or two people answering). Not a multi-tenant product. Stack: Next.js admin, Supabase (Postgres, auth, Realtime), and a small embeddable widget script that renders the chat in an iframe. Data model: contacts (id, external_user_id, email, name, verified), conversations (id, contact_id, status: open | closed, assignee_id, last_message_at), messages (id, conversation_id, sender: contact | agent | bot, body, created_at, read_at), articles (slug, title, body_md). Requirements: - Widget: open bubble, send messages, see agent replies live, show an email field when no agent is online. - Admin inbox with conversation list, unread counts, close and reopen. - When an agent replies and the visitor is offline, email the reply; inbound replies to that email thread back into the conversation. - Optional: suggest an answer from help articles using an LLM, shown to the agent before sending. Security: - Identify logged-in customers with an HMAC of their user ID computed on my server; reject unsigned identities. - Render all messages as plain text to prevent XSS in both widget and inbox. - Rate limit anonymous messages per IP and add a honeypot field. - Row Level Security so a visitor token can only read its own conversation.
Security pitfalls to check
AI tools often get these wrong by default. Check each one before anyone else uses the app.
- 01Without identity verification, anyone can open the widget and claim to be a customer by passing their email or user ID.
- 02Chat messages rendered as HTML create stored XSS in the agent inbox, which is a high-value target.
- 03Anonymous conversation tokens must only grant access to that conversation; enforce it with Row Level Security.
- 04Public chat endpoints attract spam and abuse; rate limit and add bot protection.
- 05AI answers can leak internal notes if the retrieval index includes private data; index only public help content.
Indie makers who built a real support business
They went far past the prototype: support, payments, reliability and years of iteration. Their profiles show what that took.
$864K MRRYasser Elsaid built Chatbase
AI agents for customer support, reported at $864K MRR.
Bhanu Teja P built SiteGPT
AI chatbot trained on your website content.
Simon Høiberg built Aidbase
AI customer support with chatbot, knowledge base and ticketing.
Antoine Minoux built Fernand
Support platform for SaaS teams with inbox, knowledge base and AI agent.
Or just use these indie-made tools instead
If you need Intercom's job done rather than a project, these are built and supported by independent makers.
Chatbase
chatbase.co
Platform to build and deploy AI agents for customer support, sales and product guidance across chat, email and voice.
SiteGPT
sitegpt.ai
AI chatbot trained on your website content to answer customer questions and support tickets.
by Bhanu Teja P
Aidbase
aidbase.ai
AI customer support with chatbot, knowledge base, AI-assisted ticketing and email support.
by Simon Høiberg
Chat.sh
chat.sh
Help center that answers customer questions directly, citing the pages it used.
by Damon Chen
Fernand
getfernand.com
Customer support platform for SaaS teams with multichannel inbox, knowledge base and AI agent.
by Antoine Minoux
What building it teaches you
You learn realtime messaging, embeddable widgets, HMAC identity verification and email threading.
Can you vibe code these too?
Vibe coding Intercom: FAQ
Can you vibe code an Intercom clone?
Hard. A basic chat widget for your own site is a weekend project; a reliable multi-channel inbox with AI answers and mobile notifications is not.
How long does it take to vibe code an Intercom clone?
A working personal or internal MVP takes about 8 h–3 days with a tool like Lovable. A polished version you would rely on every day takes about 10–25 days, mostly spent on never losing a message: offline agents, email fallback and reply-by-email threading.
How much does it cost to run your own Intercom alternative?
Expect $0–$60/mo at personal or small-team scale (hosting and database with realtime, email in and out, llm api for ai answers). Intercom Essential costs $19/mo/seat (checked 2026-11-03), so the most you would save is $228 a year, before counting your time.
How hard is it to build a live chat widget?
A basic widget with realtime messages and an agent inbox takes a day or two with a realtime database. Reliability features like email fallback, identity verification and spam protection take much longer.
Can I build my own AI support chatbot instead of Intercom Fin?
Yes, retrieval over your help articles plus an LLM API is a common vibe coding project. The hard part is making it decline gracefully and hand off to a human when it does not know.
What are cheaper Intercom alternatives for small teams?
Several indie-made support and AI chat tools cost far less than per-seat Intercom plans, including Chatbase, SiteGPT and Aidbase. Pick based on whether you need live chat, AI answers or a full help desk.
Is it legal to build your own Intercom clone?
Building your own tool that solves the same problem is common and generally fine. Copying Intercom's code, design assets, brand name or content is not. Use your own name and design, respect their terms of service, and get proper advice if you plan to sell it. This page is not legal advice.