Website chat and messaging apps feed one team inbox
Chatwoot is a customer support platform. Messages from the website chat widget, email, Facebook, Instagram, Twitter, WhatsApp, Telegram, Line and SMS land in a single inbox, where the team splits up the conversations. It is a Ruby on Rails app with a Vue front end, published by Chatwoot Inc., which also sells a hosted cloud version.
Agent-only notes and a public help center
- Private notes and @mentions: agents talk about a customer inside the conversation thread, and the customer never sees it.
- Auto-assignment: conversations go to whichever agents are available, with business hours and auto-responders.
- Canned responses, keyboard shortcuts and a command bar for getting through repeat questions quickly.
- Help Center portal: articles, FAQs and guides are published from Chatwoot so customers can find answers on their own.
- Integrations: reply from Slack, see a customer's Shopify orders, create Linear tickets, plug in a Dialogflow bot, translate incoming messages with Google Translate.
- Reports: per conversation, agent, inbox, label and team, plus CSAT scores, all downloadable.
Community edition without Captain or SSO, on a four-service stack
The README's "open-source alternative to Intercom, Zendesk, Salesforce Service Cloud" is open core. The LICENSE file puts the whole repository under MIT except the enterprise/ directory, which is governed by the Chatwoot Enterprise license. That code may be copied and modified for development and testing, but production use requires a subscription for the right number of seats. Captain, SLA policies, SAML settings and custom roles all live in that directory.
The free Community edition has neither Captain nor SSO. The self-hosted pricing page charges $19 per agent per month for Premium Support, which adds Captain, voice calls, custom branding, agent capacity, and roles and permissions. SAML SSO and SLA policies require the Enterprise edition, at $99 per agent per month. Yet the README lists Captain and agent capacity management among its features without mentioning a price. Captain runs on OpenAI's GPT models, through the key and endpoint configured on the instance (CAPTAIN_OPEN_AI_API_KEY, CAPTAIN_OPEN_AI_ENDPOINT), and Captain V2 defaults to gpt-5.2.
A self-hosted instance runs four services. The repository's docker-compose.production.yaml starts the Rails app, a Sidekiq worker, PostgreSQL (the pgvector/pgvector:pg16 image) and Redis. The self-hosting documentation sets a floor of 2 cores, 4 GB of RAM and 20 GB of SSD, and recommends 4+ cores, 8 GB+, PostgreSQL 12+ and Redis 6+ in production. LibreDesk, a helpdesk shipped as a single Go binary, covers email, web chat and WhatsApp, and also needs PostgreSQL and Redis.
A new minor version ships every month. Twelve came out in twelve months, from v4.8.0 (November 19, 2025) to v4.19.0 (October 20, 2026), and the latest release is v4.19.1, published October 24, 2026. Day-to-day work lands on the default develop branch, and the README points to master or tags for a stable build. Version 4.19.0 retired Captain V1: a self-hosted install that relies on V1 instructions has to migrate them to V2 before upgrading, using a task shipped since v4.16.0 and described in the Captain V1 retirement guide.