Chatty
Open-source, self-hostable AI customer support platform with chat widget, shared inbox, voice agent, RAG knowledge base, and an MCP server.
At a Glance
Free open-source software (MIT) self-hosted on your own infrastructure; you supply hosting, Supabase or Docker with Postgres/pgvector, and your own model API keys.
Full-featured 14-day free trial on every cloud plan, $0.00 charged today.
Engagement
Available On
Alternatives
Listed Nov 2026
About Chatty
Chatty is an AI customer support platform from PersonaliAI. It indexes a website, files, and product catalog so an AI agent can answer visitor questions, capture leads, book meetings, and talk over voice. It is available as a hosted cloud service and as an MIT-licensed open-source stack that can be self-hosted. The latest GitHub release listed is v2.1.0.
What It Is
Chatty is a customer support and conversational AI platform positioned by its README as a self-hostable Zendesk and Intercom alternative. Operators create one or more AI agents, train them on sources such as website pages, PDFs, DOCX, CSV, and text, and embed the resulting chat widget on a site with a script tag. A shared inbox lets human agents take over conversations from the AI.
Knowledge retrieval uses PostgreSQL pgvector with HNSW indexing and hybrid lexical/vector search. Agents can call tools to qualify leads, check calendars, and schedule meetings through Google Meet, Microsoft Teams, or Zoom.
Channels, Flows, and Voice
The README documents an omnichannel hub covering WhatsApp, Messenger, Instagram Direct, TikTok Business Messaging, Telegram, Twilio SMS, and inbound support email, all with human takeover. A visual flow builder, built on @xyflow/react, supports trigger, message, condition, lead capture, meeting booking, webhook, and AI fallback nodes, including time-of-day branching.
A LiveKit-based voice agent supports low-latency bidirectional voice, with modular STT/LLM/TTS pipelines or multimodal realtime models, five audio visualizers, and embed options for web, React, iOS, Android, Flutter, and React Native. WooCommerce sync allows order status and product inventory lookups.
MCP Server and Agent Control
Chatty ships a hosted MCP server with 81 tools across areas such as bot lifecycle, widget design, flows, campaigns, knowledge base, help center, inbox, leads, analytics, WooCommerce, security, and voice. Authentication uses OAuth 2.0 with PKCE and dynamic client registration, plus API keys. The site documents connecting from ChatGPT, Claude, Claude Code, Cursor, Gemini, Codex, Windsurf, and others, and a Codex plugin marketplace source is published in the GitHub repo.
Self-Hosting and Privacy
The self-hosted deployment runs as two Docker services (FastAPI backend and Next.js frontend) backed by a managed Supabase project for Auth, Postgres, Storage, and Realtime. Render and Railway deploy buttons and a Render blueprint are included. Models can be supplied via bring-your-own-key for Gemini, OpenAI, Anthropic, Groq, or OpenRouter. The project states GDPR-oriented controls including automated PII scrubbing, visitor right to erasure, and no training of foundation models on customer data.
Community Discussions
Be the first to start a conversation about Chatty
Share your experience with Chatty, ask questions, or help others learn from your insights.
Pricing
Self-Hosted Open Source
Free open-source software (MIT) self-hosted on your own infrastructure; you supply hosting, Supabase or Docker with Postgres/pgvector, and your own model API keys.
- MIT licensed
- Self-host on your own server
- Bring your own model API keys (Gemini, OpenAI, Anthropic, OpenRouter)
- Zero software licensing costs
Free trial
Full-featured 14-day free trial on every cloud plan, $0.00 charged today.
- Full-featured
- Cancel anytime
Hobby
Perfect for individuals, developers, and side projects.
- 1,000 AI credits/mo (~800 conversations)
- 10M training characters
- 500 MB Media Storage (250 products)
- 3 AI agents
- Fast & Advanced AI models
- AI Actions & Analytics
- Guardrails & Notifications
- Lead collection & API
- Provider-managed AI routing
Standard
All in Hobby, plus advanced automation and multi-bot systems.
- 10,000 AI credits/mo (~8,000 conversations)
- 20M training characters
- 2 GB Media Storage (1,500 products)
- 6 AI agents
- Daily Auto Train sync
- Remove branding completely
- Unlimited team members
Business
For enterprise scale, heavy traffic, and reseller options.
- 40,000 AI credits/mo (~35,000 conversations)
- 50M training characters
- 10 GB Media Storage (Unlimited products)
- Unlimited agent configurations (fair-use limits apply)
- BYOK (Bring-Your-Own-Key) option
- White-label configuration
- Management Admin API
Capabilities
Key Features
- Website, file, and catalog knowledge training with RAG
- Embeddable chat widget with streaming replies
- Shared inbox with live human takeover
- Omnichannel messaging: WhatsApp, Messenger, Instagram, TikTok, Telegram, SMS, email
- Visual conversation flow builder
- LiveKit WebRTC voice agent
- Lead capture and meeting booking
- WooCommerce order and product sync
- Proactive campaigns with behavioral triggers
- Analytics and performance scorecard
- Guardrails, domain allow list, and PII scrubbing
- Bring-your-own-key model support
- 81-tool MCP server with OAuth 2.0 PKCE
- Help center article publishing
- White-label and management API
- Auto train sync for knowledge sources
