They had one question
Does it cover my case? Is there a refund? Is this the right plan? One sentence would have settled it.
One question sits between a visitor and a sale. Salut shows you what they are reading, writes the answer, and sends it before they close the tab.
The silent drop-off
Your best prospects are the ones who almost bought. They are also the ones you never hear from.
Does it cover my case? Is there a refund? Is this the right plan? One sentence would have settled it.
Writing to a stranger and waiting a day is more effort than closing the tab. So they close the tab.
Analytics tells you someone left the pricing page. It never tells you what stopped them.
A chat bubble on the page turns that silence into a sentence. Salut is built so answering it takes you fifteen seconds, not fifteen minutes.
How it works
Other chat tools hand you a message in a vacuum. Salut hands you the message and everything around it.
The page they are on, how long they have been there, where they came from, and who they are if they are signed in. No more asking which plan they mean.
Salut reads your site and the context you wrote yourself, then prepares a reply. You approve it, edit it, or ignore it and write your own.
One message, in context, while they are still on the page. The conversation closes and the inbox goes back to empty.
Free covers 1 site, so you are good. Pro adds volume and your own AI account.
Clean conversations
Support tools grow into a second job. Salut is built around getting back to zero, so the chat never becomes the thing you avoid opening.
Open, in progress, closed. No tags to maintain, no queues to design, no workflow to configure before the first reply.
Questions you have answered a hundred times get handled from the context you wrote once. What is left is the conversation that actually needs you.
Sweep resolved threads in one action from the dashboard, the API, or your agent. The inbox is meant to be empty.
Built for agents
Salut was designed to be driven by software as comfortably as by a person. Everything you can do in the dashboard, an agent can do over the API, including the install.
About thirty tools over one endpoint: conversations, messages, websites, AI context, memories, notifications and analytics. Authenticated with an sl_ key you create in the dashboard.
POST /api/v1/tools/{tool}
The same tools, spoken natively by your assistant. Ready-made config for Claude Code, Claude Desktop and Cursor, plus any other MCP client you already run.
https://salut.chat/api/mcp
The moment a visitor sends a message, Salut posts it to your endpoint, signed with HMAC-SHA256 so you can verify it came from us before you act on it.
X-Salut-Signature: sha256=…
Give the Salut support AI an endpoint of yours, such as order lookup, licence check or seat count, and it calls it live inside the conversation. Visitor identity is injected server-side, never by the model.
POST /hooks/order-lookup
There is a written brief in the docs your agent can follow end to end: create the website, write the AI context, install the widget, verify it fired. You paste one prompt and check the result.
Is the free plan enough if I only run one site?
Yes. Free covers 1 site, you are set.
11:23On your phone
You are not sitting at a helpdesk. The iOS app exists so the two minutes that decide a sale do not have to wait until you are back at your laptop.
A push the second someone writes, not a digest tomorrow morning. Reply from the notification and the visitor sees it before they leave the page.
It answers on its own and taps you only when it cannot. Take over from the train and the conversation is closed by the next stop.
The full conversation history, visitor context and status changes. Not a viewer, the actual inbox.
Setup
Four steps, no configuration meeting, no implementation call. Every step ships with a prompt you can paste straight into your coding agent.
One field. Salut creates the website, reads your public pages, and writes the first draft of the context its AI will answer from.
Copy one script tag, or copy the ready-written prompt and hand it to Claude Code, Cursor or your own agent. It installs itself and reports back.
Deploy whenever you like. Salut watches for the widget to come online and tells you the moment it does. Skip the wait if you already know it is live.
Open your own site, say hello, and watch it arrive. When you see it land you are done: that is the whole setup.
Pricing
No seat minimum, no onboarding fee, no annual contract to sign before you have had your first conversation.
Start taking visitor questions today.
$0forever
One site, your own AI account, room to grow.
$20per month
or $200 a year
Several sites and a full team in one inbox.
$99per month
or $990 a year
The API and the MCP server work on every plan. Go past a monthly number on a paid plan and Salut buys what the next message needs: 100 more AI replies for $3, 500 more conversations for $2, up to a spend cap you set, $50 by default. At the cap chat keeps running and only the AI pauses. Connect your own model account and its replies stop counting at all.
Yes, that is the default. A new website replies on its own from the context you write during onboarding, and hands the thread to you when it cannot answer. Turn auto-replies off per website whenever you would rather write every reply yourself.
One script tag before the closing body tag. If you use a coding agent, onboarding hands you a written prompt and the agent does the edit for you.
Yes. If your visitor is signed in, pass their details to the widget with a server-computed HMAC hash and Salut marks the identity as verified. Nobody can impersonate a customer by editing their browser console.
A bubble in the corner with your colour and your name. It loads asynchronously, so it never blocks your page, and it works on mobile Safari.
That is the point. Every dashboard action is available as a tool over the REST API and the MCP server: create a website, write the AI context, read conversations, send messages, close threads, pull stats.
Chat keeps working: visitors still reach you and you still answer by hand. Only the AI pauses, and new conversations hand off to your team instead. We warn you at 80% and email you the moment it happens, and everything resumes on your next billing date.
A visitor writing to you for the first time. Opening the widget and leaving is not a conversation, spam is not a conversation, and every later message in the same thread is free. You are billed once, when someone actually starts talking.
Yes, on any paid plan. Connect your model account or paste your own API key, pick any model that key can reach, and Salut runs replies on your account instead of ours. Those replies never count against your monthly AI allowance.
No. The free plan covers one website and never expires. Pro and Business add a 14-day trial on top of that if you want the bigger allowances.
Yes: email, and a signed webhook that fires the moment a visitor writes, so you can route it into Slack, Telegram or anything else you already run.