Designed around the way small businesses actually communicate — across channels, on their own schedule, in their own voice.
Training
No data science degree. No prompt engineering. Five surfaces, all owner-editable, all in the dashboard.
One click. We read every page, chunk it, embed it locally on UK GPU hardware. Re-crawl when you publish updates.
Add anything that lives in graphics on your site (price cards, package artwork) — the bot quotes them verbatim.
Common questions + your preferred answers. The bot uses them almost word-for-word when matched.
One short tone note + a handful of past replies. The bot mirrors your style — warmly, formally, however you actually write.
Channels
One brain, four channels. The same training answers everywhere — consistently.
One line of HTML on any site (WordPress, Shopify, Wix, Squarespace, Webflow, plain HTML). Themable colour + greeting.
Forward customer enquiries to a MeshX address. We draft a reply, route it through your own Resend domain or ours.
Pair your existing WhatsApp number once (QR code, like WhatsApp Web). Replies go from YOUR number. Your phone doesn't need to stay open.
Connect your Facebook Page. The bot answers DMs the same way it answers website chats.
Control
The bot is a colleague, not a replacement. Built-in guardrails put you firmly in charge.
Start in Draft: every reply waits for your approval in the inbox. Switch to Autonomous when confidence is high.
Click cells on a weekly grid: only nights, only weekends, only working hours. Or always-on. Outside the schedule? Escalates to you.
Set how strict you want the bot to be. Replies below the threshold automatically escalate instead of being sent.
Refunds, complaints, legal, medical — automatically handed back to a human. Never auto-handled, no matter how high confidence.
Inbox
The bot handles the routine; you handle the nuance. The inbox makes that handoff seamless.
Every conversation — web widget, email, WhatsApp, Messenger — lives in one timeline.
See the bot's proposed reply with its confidence score. Approve, edit, or discard. Or reply yourself from the same page.
When the bot escalates, it tells you why — outside knowledge base, complaint keyword, low confidence, etc.
Privacy & performance
Inference runs on UK GPU machines we own. No OpenAI / Anthropic / Google calls. No per-token bill — flat monthly fee.
Each business gets its own workspace. Row-level security at the database layer. Your training data is yours alone.
Every message is logged in your inbox. Export it whenever. Delete it whenever. We don't train on your customer conversations.
Set up takes an afternoon. First reply usually goes out the same day. No credit card needed.
Start free trial