If you're evaluating open-source alternatives to Intercom, Drift, or Zendesk Chat, you've likely come across three names: Chatwoot, Typebot, and ChatterMate. They all live in the "self-hostable customer chat" space — but they solve very different problems.
Here's an honest breakdown.
The Quick Summary
| Feature | ChatterMate | Chatwoot | Typebot |
|---|---|---|---|
| Primary focus | AI-first chatbot automation | Human agent inbox | No-code bot builder |
| Local LLM support | ✅ Ollama native | ❌ | ❌ |
| No-code workflow builder | ✅ | ❌ | ✅ |
| Human handoff | ✅ Smart escalation | ✅ Core feature | ✅ Basic |
| License | AGPL-3.0 | MIT | AGPL-3.0 |
| Self-hostable | ✅ Docker | ✅ Docker | ✅ Docker |
| Vector search / RAG | ✅ pgvector | ❌ | ❌ |
| API-first | ✅ FastAPI | ✅ Rails API | ✅ |
| Multi-LLM provider | ✅ (OpenAI, Groq, Ollama, Google) | ❌ | ❌ |
Chatwoot: The Human Agent Inbox
Chatwoot is excellent at what it does: routing customer conversations to human agents. It's battle-tested, mature, and has a large community.
Best for: Teams with human support agents who need a shared inbox, SLA tracking, and omnichannel support (email, Twitter/X, WhatsApp, etc.)
Not ideal for: AI-first automation where you want bots handling the majority of conversations without human involvement.
If you want AI, you'd need to bolt it on externally — Chatwoot doesn't natively run language models.
Typebot: The Flow Builder
Typebot is a drag-and-drop conversation flow builder. Think visual programming for chatbots — you build branching dialogue trees, collect form data, and trigger webhooks.
Best for: Marketing teams building lead capture bots, survey flows, or onboarding sequences.
Not ideal for: Dynamic AI responses. Typebot's bots are deterministic — they follow the script you build. You can integrate OpenAI via blocks, but the LLM is a plugin, not the core.
ChatterMate: The AI-First Agent
ChatterMate takes a different approach: the AI is the agent, not a bolt-on.
You connect a knowledge base, configure an agent, and ChatterMate uses vector search (pgvector) to give the AI relevant context before responding. Customers get accurate, contextual answers 24/7 — and the agent escalates to humans when it can't help.
What makes it different:
1. Fully local AI with Ollama
Run ollama pull llama3.2 and ChatterMate connects to it directly. No OpenAI account, no API costs, no data leaving your infrastructure. This matters enormously for regulated industries (healthcare, finance, legal).
2. Multi-LLM switching
Already have OpenAI access? Use GPT-4o. Want to experiment with Groq's speed? Switch providers without changing your setup. Per-agent model selection means you can run different LLMs for different use cases.
3. Smart human handoff
ChatterMate doesn't just dump the customer to a human — it's aware of business hours, routes based on team availability, and includes conversation context so the agent isn't starting from scratch.
4. RAG-ready from day one
pgvector is baked in. Upload documents, connect to your knowledge base, and the AI has semantic search over your content before every response.
5. No-code workflow builder
For custom logic — conditional responses, Jira ticket creation, form collection — there's a drag-and-drop builder. No Python required.
# Get started in minutes
git clone https://github.com/chattermate/chattermate.chat.git
cd chattermate.chat
cp .env.example .env
docker compose up -d
When to Pick Each
Pick Chatwoot if: You primarily have human support agents who need a shared inbox, omnichannel support, or SLA tracking.
Pick Typebot if: You're building marketing or onboarding flows and want visual, no-code flow control.
Pick ChatterMate if: You want AI handling 80%+ of conversations autonomously, data sovereignty matters (fully local with Ollama), or you need smart human escalation baked in.
The Honest Take
These tools aren't really competing — they're solving different problems. If you want a chatbot that thinks, ChatterMate is the open-source option built for it.
⭐ GitHub: chattermate/chattermate.chat — 58 stars, AGPL licensed, v1.0.9
Contributions welcome — zero open issues means the board is yours.
I'm Arun, building ChatterMate in public. Follow @ChatterMat65796 for updates.
