One brain across all your channels
Set it up once and it answers the same on WhatsApp, Instagram, your web and your app.
- Web widget
- iOS SDK
- Android SDK
- Flutter SDK
It answers your users, looks up their data and acts through your app’s APIs
Copy the command and paste it in your terminal to add Vatio to Claude Code.
Built and kept up to date by the coding agent you already use. See how
Set it up once and it answers the same on WhatsApp, Instagram, your web and your app.
It recognizes your signed-in users, and their data only opens with that verified session.
sub: user_8f2c GET /api/orders/4821
Authorization: Bearer <Sofía’s token>
It looks up orders, bookings or accounts in your system instead of making up an answer.
| Tool | Calls | OK |
|---|---|---|
get_orderprivate | 312 | 100% |
list_bookingsprivate | 128 | 99% |
reschedule_bookingprivate | 41 | 88% |
notify_team | 6 | 100% |
Your team writes the policies and answers, and the agent replies with those, not guesses.
Knowledge9 entries
When a case calls for it, it hands the chat to your team and stays quiet while they help.
I want a refund for #4821
Refunds are handled by our team. I’ve let them know, and someone will reply here.
Marta took over
Hi Tom, I’m Marta. Your refund is on its way.
You flag a wrong reply and it becomes a test that runs on every version.
What it should have done
Look the order up with get_order before saying it doesn’t exist.
Test case added · replayed on every version
Your team gets each change at once, your users once the tests that pass today still pass. A change that breaks one is held back.
Undo the last change
Vatio checks every reply before sending it and fixes the ones with links or text that shouldn’t be there.
Grab the discount at acme-deals.comReply discarded · unauthorized link
Two-step verification, roles for your team and a log of every change and every access to your conversations.
Two-step verificationActivity log
Three pieces, so the work happens in your project, with the tools it already uses.
MCP
vatio.ai/mcp Build, test, change and supervise your agent from your coding agent or any chat. No config files in your repo.
Skills
npx @vatio-ai/skills The whole workflow, taught: /vatio-init builds and tests, /vatio-improve fixes.
Widget and SDKs
@vatio-ai/widget The agent in your app: one script tag on the web, a package on iOS, Android and Flutter.
When you build your third agent that talks to users, you realize you’ve already reimplemented a lot of things that should just work by default: channel connections, a knowledge base, safeguards, conversation management and more. All of it is needed regardless of the agent’s personality, goal and tools.
Vatio is the infrastructure and runtime the agent you’re building needs to answer real users safely.
Vatio is also built for your coding agent. With our MCP, your coding agent can build, launch, test and improve your agent quickly and efficiently, and there are no config files to keep in your repo.
If you need a reliable agent for your product, try Vatio and let your coding agent do the hard work.
Your coding agent builds the agent. Your plan covers running it on Vatio.
For your first users.
US$0CLP 0/month
For more conversations.
US$49CLP 49,000/month
or US$490CLP 490,000 a year
Save US$98CLP 98,000: two months free
Pro activation is assisted after signup. No automatic charge. Request activation →
For apps with many users.
US$149CLP 149,000/month
or US$1,490CLP 1,490,000 a year
Save US$298CLP 298,000: two months free
Business activation is assisted after signup. No automatic charge. Request activation →
US dollars, excluding taxes.Chilean pesos, final price.
No. You tell your coding agent (Claude Code, Cursor or Codex) what your users need, and it builds the agent and connects it to your app. You try it, decide what it may do and approve each change. You only need your app’s project on your computer and Node.js 20 or newer.
Yes, once its code is on your computer: connect the project to GitHub, open it in Claude Code or Cursor and run the second line there. The first one works from any chat.
On Vatio. Your coding agent builds and updates it from your project; Vatio runs the conversations. Start in your app or site, and add WhatsApp or Instagram when you need them.
It can get things wrong, like any agent. It answers with the knowledge your team writes and the data from your system, and Vatio checks every reply before sending it: it fixes empty ones and ones with links you never gave it. The tests cover the mistakes you report; they can’t guarantee every possible answer is right.
Yes. Every conversation shows up in the inbox. When a case calls for it, it hands the chat to a person on your team and the agent stays quiet while they help.
Only what you give it through tools that call your API. A user’s data only opens with their verified session, and your system decides what they can see. Your keys are never visible to the agent.
Conversations stay in your Vatio account, and every access to them is logged. The models Vatio uses don’t keep what your customers write.
Yes, besides chat in your app or site. To try it you use a shared number and account, with no paperwork with Meta; you connect yours when you want to serve real customers, on any plan.
In part. You or your team flag wrong replies in the inbox, and each one becomes a test that runs on every new version. Your coding agent prepares the fixes you approve, and each one reaches your users only if nothing that worked breaks.
Yes. You sign in with your email, Google or GitHub, and Free includes web chat and 100 conversations a month. A conversation is a chat with a real person, counted once a month; tests don’t count. If you need more, the Vatio team activates Pro or Business after signup.
Copy the command and paste it in your terminal to add Vatio to Claude Code.