One runtime for the web, chat apps, voice, and devices.
Assistant Core brings identity, prompt, models, voice, knowledge, memory, tools and MCP, messaging channels, and the device gateway into one runtime, so an assistant behaves the same on every channel.
- Website, chat apps & browser voice
- Developer API, WebSocket & MQTT
- Knowledge, memory, tools & MCP
- LLM providers, switch per assistant
- 6
- Messaging channels on the same runtime
- 6
- To first audio in Realtime voice
- ~0.7s
What the runtime covers
Instead of wiring disconnected services together, product teams get one assistant runtime that can be configured, operated, and extended per tenant.
Assistant intelligence
Each assistant has its own identity, system prompt, model, knowledge base, memory, and domain.
- Per-assistant prompt and persona
- Answers grounded in your private knowledge
- Long-term memory, with long chats summarized to stay in context
Voice runtime
Two voice modes, chosen per assistant, with the heavy work in the cloud instead of on the device.
- Realtime speech-to-speech for the fastest replies
- Standard ASR → LLM → TTS pipeline you tune per assistant
- Natural interruptions in both modes
Tools, MCP & scheduled tasks
Assistants call built-in tools, your REST APIs, MCP servers, or device-side tools, and can run tasks on a schedule.
- Tools enabled per assistant and per channel
- MCP servers, MCP endpoint, and device MCP
- Scheduled tasks with results in their own conversation
Integration surfaces
The same assistant appears on your website, in messaging apps, in your own apps, by voice, and on hardware.
- Website widget and Developer API (SSE chat, voice sessions)
- Zalo OA, Messenger, Instagram, WhatsApp, Telegram, TikTok Shop
- MQTT gateway for smart devices
Operations layer
Manage assistants, users, roles, channels, devices, conversations, and credit usage in one admin panel.
- Per-assistant roles and permissions
- Conversation history, feedback, and LLM traces
- Credit usage and prompt cache hit rate
Multi-tenant foundation
Each assistant is isolated by domain, configuration, data, and access control.
- Request-level tenant isolation
- Own subdomain or custom domain and branding
- Assistant context cached by domain
From one assistant to every channel
Start on one channel, then add voice, messaging apps, devices, and scheduled tasks without rebuilding the assistant.
- 01
Create the assistant
Set its name, domain, system prompt, model, and voice.
- 02
Add knowledge and memory
Upload documents or crawl websites so answers stay grounded, and let memory keep each user's context.
- 03
Connect tools and MCP
Let the assistant call your APIs, MCP servers, or tools exposed by connected devices.
- 04
Ship to every channel
Embed the website widget, connect Zalo OA and other messaging apps, use the Developer API, or pair ESP32 hardware.
Runtime FAQ
Explore next
How is this different from a chatbot UI?
A chatbot UI covers the conversation screen. Assistant Core runs everything behind it: prompt, models, voice, memory, knowledge base, tools and MCP, messaging channels, device connections, roles, and traces.
Do smart devices need to run a local model?
No. The device keeps the microphone, speaker, sensors, and actuators, while the cloud runtime handles speech, reasoning, memory, tools, and voice output.
Can one assistant be shared across the website, chat apps, and devices?
Yes. That is the core idea: one assistant shares its prompt, knowledge, memory, and tools across the website widget, messaging channels, browser voice, and smart devices.
Hear it before you build it.
Talk to an Assistant Core assistant right now, then create your own in a few minutes.
Free plan available.