First token in under 300ms
Streaming starts before a human could read the message. AI SDK v6 on Cloudflare Workers edge. Groq on hot paths drops below 150ms. No cold starts.
Multi-channel AI
One brain. Every channel.
Web, WhatsApp, Telegram, Discord, iMessage — same memory, same agent. First token in under 300ms.
60 seconds to a live chatbot · No code required
Split view: same customer conversation on WhatsApp (9pm) and web chat (the following morning) — one thread, one memory
<300ms
First token (streaming starts)
5
Channels live from one agent definition
60s
From blank markdown to deployed chatbot
53%
Mobile users leave sites >3s — Google 2017
"847 messages in the first week. We handled 94% without a human. The 6% that reached us were exactly the right ones."
Live on every channel in four steps
1
Write the agent
A markdown file: frontmatter sets the model, system prompt, skills, tools, and channels. Takes minutes in any text editor or IDE.
agents/*.md · any editor · no special syntax
2
Pick the channels
Toggle web, WhatsApp, Telegram, Discord, iMessage in the workspace dashboard. One toggle per channel. Active immediately.
Dashboard → Channels → toggle on/off
3
Publish
`oneie agent publish` — the agent is live on every selected channel simultaneously. Deployment takes under 60 seconds.
oneie agent publish · <60 seconds
4
It remembers
Every message links to an actor record. The same customer across channels is one record. History loads before every response.
actor consolidation · TypeDB · D1 transcripts
What the agent can do
One agent definition. Six outcomes.
Generative UI on web
The web channel renders real components mid-stream: payment cards, booking forms, maps, product grids. Not just text — actual interactive UI built by the agent.
Cross-channel memory
Same customer on WhatsApp today = same actor from their Discord message last week. One thread, one history. The agent knows who it's talking to across every channel.
5 channels, 1 agent definition
Web · WhatsApp · Telegram · Discord · iMessage. One markdown file defines the agent. Channel adapters handle the rest. Toggle channels per client workspace.
Agent teams with handoffs
Receptionist → Sales → Manager. Complex queries escalate automatically. Every handoff is a signal. Every resolution marks the path. The routing gets smarter every week.
Group chats and @mentions
Discord servers and Telegram groups are supported. The agent responds when @mentioned or when configured keywords match. Groups get a dedicated inbox in the agency dashboard.
How we compare
| Feature | ONE | Intercom | ManyChat |
|---|---|---|---|
| WhatsApp native | |||
| iMessage (Business Chat) | |||
| Cross-channel memory (one actor) | |||
| Generative UI on web | |||
| First token <300ms | |||
| Per-client learning | |||
| White-label brand | |||
| Mature helpdesk workflow (large support teams)Honesty: Intercom wins here |
Credit-based pricing
Pay for what your agents actually do. Annual billing pre-selected.
Starter
3 channels · 2 agents
$500/mo
5M credits/mo
Web + Telegram + WhatsApp
2 agent workspaces
<300ms first token
Basic cross-channel memory
Most popular
Agency
All 5 channels · unlimited agents
$5,000/mo
60M credits/mo
All 5 channels (+ iMessage + Discord)
Unlimited agents + clients
Generative UI on web
Group chat + @mention routing
Priority edge routing
Scale
Dedicated edge · SLA 99.99%
$50,000
750M credits/mo
Dedicated edge region
SLA 99.99%
Custom channel adapters
<150ms first token guarantee
Groq reserved capacity
Questions about channels
Your clients' customers are waiting on every channel.
One agent definition. Five channels. 300ms first token. Start the chatbot that never sleeps.
60 seconds to a live chatbot · No code required