Customer support · checked 2026-10-19Can I vibecode Chatwoot Cloud?
KINDA· weekend project
A chat bubble and a shared inbox for one website is a weekend: a widget script, a WebSocket server, an agent page, and SMTP plus IMAP for email. The gaps open after that: inbound mail that threads correctly, spam, WhatsApp and Instagram with their API approvals, and a mobile app for agents. Chatwoot's code is open under a mostly MIT licence and runs from Docker, so for most people the cheaper route is to host the same software on a small server and skip Cloud's per-agent fee. A build from scratch only makes sense if you want something much smaller.
$19/moprice per seat
$228you'd save a year
weekendbuild time
—people replaced it
What you need on hand.
DifficultyIntermediate · Some system setup or keys.
TimeA weekend; email threading and the reverse proxy take the most care
- Claude Code, Codex or CursorAny of them can run the prompt.
- Node.js 22better-sqlite3 compiles a native module, so build tools must be installed.
- A support mailbox with SMTP and IMAPAn app password works; keep it off your personal inbox.
- A server with HTTPSA small VPS with Caddy or nginx; the proxy must pass WebSocket upgrades.
Before you buildChatwoot's Hacker plan is free for two agents with live chat, and the Community Edition is free to self-host. Try either before building.
- 1. Install one agent: Claude Code, Codex, Cursor.
- 2. Make an empty folder and open it in the agent.
- 3. Paste the prompt below, press enter, and answer its questions as it builds.
The prompt
Build me a small support desk for my site to replace Chatwoot Cloud. Requirements: - Node.js 22 + TypeScript: a Fastify server with `@fastify/websocket` and a React inbox built with Vite. `npm start` serves http://localhost:4760: the inbox at /app and the widget at /widget.js. - Storage: better-sqlite3 in ./data/desk.db: contacts (email, name, custom attributes as JSON), inboxes (website widget, support email address), conversations (open, pending, snoozed, resolved; assignee; labels), messages (incoming, outgoing, private note, attachment paths), canned responses, agents. - Agents: up to 5, created with `npm run agent:add`, password hashed with argon2id, session cookie. Roles are admin and agent; only an admin edits inboxes, labels and canned responses. - Widget: one script tag shows a bubble in a Shadow DOM. The visitor id lives in localStorage, a pre-chat form asks for name and email, a dropped connection reconnects and re-sends unconfirmed messages, and outside business hours it shows an away message. - Email channel: poll IMAP with `imapflow` and thread on Message-ID, In-Reply-To and References. Replies go out through nodemailer with the same headers. Save attachments under ./data/files (10 MB each), sanitize HTML mail with `sanitize-html` and never load remote images. - Inbox: lists filtered by mine, unassigned, status and label; assign, label, add private notes, type "/" for canned responses; a side panel with the contact and its past conversations; resolve conversations idle for N days (setting). - Security: the widget only works from ALLOWED_ORIGINS, 20 messages a minute per visitor, 2,000 characters per message, everything escaped on output, secrets in .env. - Single team. No visitor accounts, no telemetry, local-first. Outbound calls: SMTP and IMAP. - Out of scope: WhatsApp, Instagram, SMS and voice channels, an AI assistant, mobile apps, SLAs, automation rules. - README: the embed snippet, the .env keys, a reverse proxy with WebSocket upgrade headers, SPF and DKIM for the support address, and a note that self-hosting Chatwoot already offers all of this.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
Then keep going
Next prompts, same folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Chatwoot Cloud.
Add a Chatwoot importer to the support desk in this folder. Requirements: - A command `desk import-chatwoot` that takes CHATWOOT_URL, CHATWOOT_ACCOUNT_ID and CHATWOOT_TOKEN (a personal access token) from .env. - Page through GET /api/v1/accounts/{id}/contacts and /conversations with the api_access_token header until a page comes back empty (check the current API reference for paging parameters), then fetch each conversation's messages. - Map contacts by email, conversations to the closest status, and messages to incoming, outgoing or private note; keep original timestamps and sender names. - Skip anything already imported, using the Chatwoot id as a stored external key, and print counts. - Update the README with the rate limit I should respect and how to find the token.Add reply drafts to the support desk in this folder. Requirements: - Index the Markdown files in ./kb at startup (split by heading, embed with an OpenAI-compatible embeddings endpoint from .env, store vectors in SQLite). - A "Suggest reply" button in the inbox retrieves the 3 best-matching chunks for the latest customer message, asks the chat model for a draft that cites the article names, and puts the draft in the reply box without sending it. - Say "no matching article" instead of inventing an answer when the best match is weak. - Log which articles were used per conversation. - Update the README with what text leaves the machine.
Add a Telegram inbox to the support desk in this folder. Requirements: - Create a Telegram channel type with a bot token in .env and long polling through `grammy`. - Each Telegram chat becomes a contact and a conversation; incoming text, photos and documents appear in the inbox, and agent replies are sent back through the bot. - Show the channel icon in the conversation list and the contact panel. - Rate-limit replies to respect Telegram's limits and queue on failure with 3 retries. - Update the README with creating the bot in BotFather.
What you lose
- WhatsApp, Instagram, Facebook, Telegram, SMS and voice channels
- The iOS and Android apps for agents
- Teams, automation rules, SLAs and the built-in reports
- The Captain AI assistant, copilot and reply suggestions
- A help center whose articles show inside the chat widget
Why people still pay
Cloud means nobody patches a server: Chatwoot runs the app, its database and Redis, the channel connectors and the mail sending, and a free tier lets two agents start at no cost. Self-hosting asks for at least 2 CPU cores and 4 GB of RAM and leaves updates, security and backups to you, which is the part a small team happily pays $19 a seat to skip.
Chatwoot Cloud or your own copy?
Seats
Over
Your usage
12h ×$/h
Chatwoot Cloud, 1 yr$228
Your own copy, 1 yr$46running + build session
You keep$182pays for itself in the first month
Estimate · On your own machine it costs nothing; a public widget needs a VPS at about $5 a month. The build is token usage on your agent plan.
Take your data, then cancel.
1 · Take your data out of Chatwoot Cloud
- 1Contacts: open the Contacts page and use Export, which gives a CSV of your contacts.
- 2Conversations and messages: use the API. Create a personal access token in your profile settings and call GET /api/v1/accounts/{account_id}/conversations on app.chatwoot.com with the token in an api_access_token header; the pricing page says conversations, contacts and other data can be exported this way at any time.
- 3Help center articles, canned responses, labels and automation rules: we found no bulk export, so copy them or pull them through the API.
2 · Cancel Chatwoot Cloud
- 1Open Settings, then Billing, and click Go to the billing portal.
- 2In the portal you can see invoices, edit billing details or cancel the subscription.
- 3The billing page then shows a Cancels on date, so the plan runs to the end of the paid period.
- 4If your plan is billed through Shopify, manage it from the Shopify link on the same page.
- The pricing page offers refunds on Cloud plans, calculated from the cancellation date and unused time. Write to sales@chatwoot.com; Captain AI credits are not refundable.
- Data older than your plan's retention (30 days on Hacker, 1 year on Startups, 2 on Business, 3 on Enterprise) is removed from the workspace, so export before you downgrade to a shorter plan.
- Moving to self-hosted Chatwoot: the install guides are public, but we found no documented Cloud to self-hosted data transfer, so plan to bring contacts in by CSV and conversations through the API.
- Keep the chat snippet on your site until the new inbox is live, then swap it in one deploy.
Checked 2026-10-19 against Chatwoot Cloud's docs: 1, 2, 3, 4, 5
The escape hatch
Don't feel like building it? These are already free.
Chatwoot (self-hosted)Open sourceThe same software as Cloud: MIT-licensed apart from an enterprise folder; runs from Docker, a Linux script or Kubernetes and needs 2 CPU cores and 4 GB RAM at least.★ 38k · active Oct 2026 · docker to self-host
Web, iOS, AndroidVisit ↗LibredeskOpen sourceAGPL-3.0 support desk shipped as a single binary, a lightweight alternative to Intercom and Zendesk.★ 3.1k · active Oct 2026 · docker to self-host
WebVisit ↗Live Helper ChatOpen sourceApache-2.0 live support with web and mobile apps, voice, video and screen sharing.★ 2.2k · active Oct 2026 · docker to self-host
Web, iOS, AndroidVisit ↗
Web, iOS, AndroidVisit ↗LibredeskOpen sourceAGPL-3.0 support desk shipped as a single binary, a lightweight alternative to Intercom and Zendesk.★ 3.1k · active Oct 2026 · docker to self-host
WebVisit ↗Live Helper ChatOpen sourceApache-2.0 live support with web and mobile apps, voice, video and screen sharing.★ 2.2k · active Oct 2026 · docker to self-host
Web, iOS, AndroidVisit ↗
The numbers
Chatwoot Cloud pricing
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Hacker | $0 | $0 | Up to 2 agents, 500 conversations a month, live chat channel only, 30 days of data retention. |
| Startups | $19 | — | Unlimited conversations (fair use), all channels, help center, 300 Captain AI credits a month, 1 year of data retention. |
| Business | $39 | — | Adds voice, teams and automation rules; 500 Captain AI credits a month; 2 years of data retention. |
| Enterprise | $99 | — | Adds SSO/SAML and audit logs; 800 Captain AI credits a month; 3 years of data retention. |
- Free tier
- Two agents, 500 conversations a month, live chat only and 30 days of retention.
- Billing
- Priced per agent per month with a 15-day free trial; the pricing page lists no annual plan. Cloud plans are refundable, calculated from the cancellation date and unused time, but Captain AI credits are not. The billing currency, US dollars or Brazilian reais, is chosen when the subscription is created and cannot be changed.
- Hidden costs
- Captain AI credits beyond the monthly allowance cost $20 per 1,000 and are non-refundable; credits you buy expire after 6 months and need an active subscription. Unlimited conversations are subject to a fair-use policy. Data older than the plan's retention is removed from the workspace and survives only in exports. Every agent seat costs the same.
- Source
- www.chatwoot.com
Also on the list
Related apps
Hacker is free for up to 2 agents with live chat only and 500 conversations a month; approved open-source projects can get Startups free on request.
Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: WhatsApp, Instagram, Facebook, Telegram, SMS and voice channels, The iOS and Android apps for agents.
About $19 a month on the Startups plan per seat, $228 a year (checked 2026-10-19).
WhatsApp, Instagram, Facebook, Telegram, SMS and voice channels; The iOS and Android apps for agents; Teams, automation rules, SLAs and the built-in reports; The Captain AI assistant, copilot and reply suggestions; A help center whose articles show inside the chat widget. If any of those matter to you, keep paying.
Yes: Chatwoot (self-hosted) (open source), Libredesk (open source), Live Helper Chat (open source).