One runtime for voice, memory, knowledge, and tools. Ship the same assistant to your website, Zalo, Messenger, apps, and ESP32 devices, built Vietnamese-first.
Free plan · No credit card · Vietnamese & English
User: Remind me to water the plants at 7 tomorrow morning. Assistant: Done! At 7:00 tomorrow I'll remind you to water the three orchids on the balcony.
Built on the models and voices you already trust
OpenAI
Anthropic
Google
xAI
DeepSeek
Xiaomi
ElevenLabs
SonioxProvider and model availability depends on deployment configuration.
Configure the assistant once. Its knowledge, memory, and tools follow it to every channel.
Connect Zalo OA, Messenger, Instagram, WhatsApp Business, Telegram, and TikTok Shop from the admin panel. Customers chat where they already are.
Zalo OA
Messenger
Instagram
WhatsApp
Telegram
TikTok ShopOne script tag, isolated in Shadow DOM. Visitors can attach images and documents when you enable it.
Realtime mode talks speech-to-speech; Standard mode runs an ASR → LLM → TTS pipeline you tune per assistant. Users can interrupt naturally in both.
Time to first audio
The device keeps the mic, speaker, and sensors. The cloud does the thinking.
Flash
Firmware for ESP32 and ESP32-S3 boards
Activate
Bind to an owner with a 6-character code
Talk
Audio over MQTT + UDP, AES-128 encrypted
Act
Sensors and controls exposed as MCP tools
Roadmap: education robots, smart toys, in-car assistants.
Device guideEvery assistant gets memory, your private knowledge, and tools, with nothing to wire together yourself.
Learns facts in the background and keeps their history. Users can view, edit, or reset what it remembers. Long conversations are summarized automatically to stay in context.
Upload documents or crawl websites. Scanned PDFs are read with OCR, then searched on every turn.
Call your APIs, MCP servers, and devices.
Ask once, and the assistant runs the task on schedule. Results land in their own conversation, with an optional email.
Pick speech-to-text, LLM, and text-to-speech per assistant, or switch to realtime. Change providers in configuration, not architecture.
Tokens, latency, tool calls, and prompt cache hit rate for every message, in the admin dashboard.
Start with the widget. Move to the streaming Chat API or Voice API when you need your own UI.
<script
src="https://cdn.assistantcore.com/embed/embed.js"
data-assistant-id="YOUR_ASSISTANT_ID"
data-api-base="https://api.assistantcore.com"
async
></script>Built and operated for Vietnamese users and products, from voice and payments to data protection.
Vietnamese speech recognition and synthesis run in the live voice pipeline, including realtime calls.
Answer customers on Zalo with the same knowledge and tools as your website.
The web app, admin panel, and documentation are available in Vietnamese and English.
Personal data handled in line with Decree 13/2023/ND-CP.
No. You create and configure an assistant in the admin panel, connect channels there, and embed the website widget with one script tag. The APIs are there when you want your own interface.
Vietnamese and English are first-class across voice, the web app, the admin panel, and the docs. The LLMs handle many more languages in text.
No. Every model, voice, and tool runs on the platform's providers and draws from one credit balance, so there are no provider keys or separate bills to manage.
ESP32 and ESP32-S3 boards running our firmware, built on the open-source xiaozhi-esp32 project. Devices stream audio to the cloud runtime, which does the reasoning.
Zalo OA, Facebook Messenger, Instagram, WhatsApp Business, Telegram, and TikTok Shop. Each connection uses the same assistant, knowledge, and tools.
Traffic is encrypted in transit, tool credentials are encrypted with KMS, access is role-based, and personal data is handled in line with Vietnam's Decree 13/2023/ND-CP. Users can view, edit, and reset their memory.
Technical deep-dives, product updates, and best practices for building AI-powered products.
Explaining Acoustic Echo Cancellation from the NLMS algorithm to real-world deployment in the browser (AEC3) and on ESP32 (ESP-ADF). Includes a comparison of AEC support across OpenAI, Gemini Live, and xAI Grok.
How we integrated OpenAI, Anthropic, Google, xAI, DeepSeek, and Xiaomi through native SDKs — letting users switch models mid-conversation, saving up to 90% on input-token costs with prompt caching, and deploying with zero downtime.
How we built a retrieval-augmented generation pipeline handling PDF, DOCX, and web URLs with chunking, embedding, and cosine search — reaching 91% accuracy compared to 34% with a plain LLM.
Talk to an Assistant Core assistant right now, then create your own in a few minutes.
Free plan available.