Webnav.ai vs Chatbase
A feature-by-feature comparison, and where Webnav.ai now stands after adding AI Actions.
TL;DR
Chatbase pioneered "AI Actions" — letting the chatbot do things (place orders, look up orders, process refunds) by calling your APIs, instead of only answering questions. Webnav.ai now ships the same capability with AI Actions, while keeping its own advantages: native live chat, native ticketing, and a self-hostable / data-stays-with-you architecture.
Feature comparison
| Capability | Chatbase | Webnav.ai |
|---|---|---|
| RAG Q&A trained on your content | ✅ | ✅ |
| Automatic website crawling | ✅ | ✅ |
| Streaming responses | ✅ | ✅ |
| Multi-language | ✅ | ✅ (en / zh / ja / ko / zht) |
| One-line embed widget | ✅ | ✅ |
| AI Actions (tool calling) | ✅ | ✅ new |
| E-commerce: place / query order, track logistics, refund | ✅ (via Shopify/Stripe + actions) | ✅ new (via your own API) |
| Confirmation step for write actions | ✅ | ✅ (server-side, exactly-once) |
| Native live chat (human agent) | ❌ (needs Zendesk/Intercom) | ✅ native WebSocket |
| Native ticket system | ❌ (needs integration) | ✅ native |
| Self-host / data stays on your infra | ❌ | ✅ |
| Lead collection | ✅ | ✅ (chat form + auto-capture + dashboard) |
| Open REST API + outbound Webhooks (Zapier / Make / CRM) | ✅ | ✅ (API keys + signed webhooks) |
| MCP (Model Context Protocol) tool servers | partial | ✅ (JSON-RPC over HTTP) |
| Multi-channel (WhatsApp / Slack / Messenger / Email) | ✅ | ✅ shipped¹ |
| Voice AI (speech in/out) | ✅ | ✅ shipped² |
¹ Channel framework + Slack/WhatsApp/Messenger adapters + a credential-free generic channel are built. Connecting a real Slack/WhatsApp workspace requires you to provide that platform's tokens in Admin → 多渠道接入.
² In-widget voice input (STT) and reply read-out (TTS) via the browser Web Speech API — enable with data-voice="true". Works in browsers that support Web Speech.
How "doing things" works in both
Neither product owns your orders or your money. Both act as an orchestration layer:
- The LLM understands intent ("where's my order?").
- It decides to call a registered action (tool) and extracts parameters.
- The platform calls your backend API with those parameters.
- Your API returns data; the LLM turns it into a natural reply.
In Webnav.ai this is the AI Actions feature, and the contract your platform implements is the E-commerce Integration spec.
Where Webnav.ai is intentionally different
- You keep the data. Webnav.ai never stores your orders, customers, or payment data. It only forwards signed requests to your API and relays the answer. Identity tokens are passed through opaquely.
- Native human handoff. Live chat and tickets are built in — no third-party help-desk subscription required.
- Open / self-hostable. Run the whole stack on your own infrastructure.
Now shipped (previously roadmap)
All five former roadmap items have landed:
- Lead collection — in-chat lead form, automatic email capture, and a Leads dashboard.
- Open API + Webhooks — Bearer API keys for
/api/v1/*plus HMAC-signed outbound webhooks; this is how Zapier / Make / your CRM connect. See Zapier / Make integration. - MCP tool servers — connect a JSON-RPC-over-HTTP MCP server as the tool backend (Admin → AI Actions, mode = MCP).
- Multi-channel — Slack / WhatsApp / Messenger adapters + a credential-free
genericchannel; configure platform tokens in Admin → 多渠道接入. - Voice AI — browser Web Speech voice input + reply read-out (
data-voice="true").