Your customers arrive with a question. Your concierge greets them, answers from your real catalog and published pages, and hands off to a person the second it’s unsure. Instant help at your front door, running on your own AI, never ours.
Your key · your model · your customers’ questionsNo scripts to write, no decision trees to wire. It reads the question, checks your real data, and then either answers or gets a human. Every time.
In plain English, in the chat on your site. No menu to navigate, no ticket to file.
Up to four quick, read-only lookups against your real catalog, prices, availability, and published pages. The same records your dashboard shows.
Every reply carries a confidence score. High enough, and it sends. Anything less, and it forks to a person.
In your customer’s hands in seconds, in your voice, at 2am and at any volume.
It tells the customer a person will follow up, then drops the conversation into your staff inbox. It would rather fetch a human than guess.
Most chat bots guess from a snapshot of your site crawled weeks ago. Yours reads the same live records your dashboard shows, the moment they change. Change a price at 9:00 and the answer is right at 9:01.
“Is the pour-over kit back in stock, and what’s your return window?”
sparx doesn’t sell you intelligence. You connect your own provider key, encrypted, and yours to revoke. Every conversation runs on your model, on your terms. This is the floor, not a setting: sparx never answers your customers on a credential of its own.
Paste your Anthropic or OpenAI key once. It’s encrypted at rest and only ever decrypted to answer.
You pay your provider directly for usage. sparx charges a flat fee, with no per-message markup.
Haven’t connected one? Every chat goes straight to a person. There is no sparx fallback AI.
Anything about a specific order, account, or refund goes straight to a person. So does any answer it isn’t sure enough about. Outside your hours it leaves your away message and passes the conversation along. Fail-safe by design.
You set your operating hours and your away message. The confidence bar is built in and set high, so anything the concierge can’t answer with certainty becomes a conversation waiting in your staff inbox.
Set it up like any other setting. Write the greeting in your voice, pick a tone, match your brand color, set your hours. No prompt engineering, no training data, no developer.
One AI module, both tools: the customer-facing concierge on this page and the agentic MCP line for your own team. Bring your own key and pay your provider for usage. Flat, on one bill, no tiers.
The concierge needs the Live Chat module too ($19/mo) — that’s the chat widget on your site and the staff inbox your team answers from. $68/mo for both, on one bill.
Your own key, where its answers come from, when a human takes over, and what it costs — answered straight. Still weighing it? Read the chat docs or start the 14-day trial.
Yes. sparx never answers your customers on an AI credential of its own — there is no house account to fall back on. You paste your own Anthropic or OpenAI key into your chat settings once, it is encrypted at rest, and every conversation runs on your model and your terms. Haven’t connected a key? The concierge stands down and every chat goes straight to your team, exactly as if AI were switched off.