Capabilities
From the first question to the next action.
Every agent we build starts from the same set of capabilities. We switch on the ones your calls need, connect them to your systems and leave the rest off.
On the call. The work your team does today.
The agent listens, understands what the caller needs and does it in your systems while they are still on the line.
Routes every call
A router picks the right workflow for each caller: orders, bookings, support or sales. Each workflow follows the script your team approved.
Works in your systems
Looks up orders, checks availability, books appointments and creates records through your own APIs. Bookings and orders are read back to the caller and sent only after they say yes.
Verifies the caller
Private records, such as an order or an account, open only after the caller proves the phone number or email on file. The match is checked in code, not by the model.
Remembers returning callers
Recognizes the number, recalls the last conversation and picks up from there, so callers do not repeat themselves.
Answers from your content
Hours, branches, policies and product details come from a knowledge base built from your own sources, not from guesswork.
Speaks the caller's language
99 languages, with natural turn-taking: it waits until the caller has finished, a little longer after a long request.
Around the call. Nothing ends at hang-up.
What the caller needs next, and what your team needs to know, goes out during the call or the moment it ends.
WhatsApp during the call
Tracking links, directions, booking details and payment pages, sent to the caller while you are still talking.
WhatsApp →Handoff to your team
When a case needs a person, the agent tells the caller what happens next and emails your team a written summary of the request.
Payments
A secure payment link from your payment provider, sent by WhatsApp during the call. Our team connects the provider during the build.
Voice payments →Outbound calls
Your systems trigger calls through the Voice API: reminders, confirmations and follow-ups, with the outcome returned to you.
Voice API →Call results
A signed event after every call with the call record; the outcome and summary follow once the call is evaluated. Transcripts and recordings are available by API unless you switch them off.
Webhooks →Integrations
Ready connectors for Shopify, Wix Bookings and monday.com. Anything else with an API, our team connects during the build.
Integrations →Guardrails in the code. Not in a prompt.
The rules that protect your callers and your data are enforced by the gateway that runs the call, so the agent cannot talk its way past them.
Security FAQ- Only the tools you approveEach agent sees the actions you switched on for it, and nothing else.
- Confirm before commitBookings and orders are read back to the caller and sent only after they say yes.
- No dead airIf a call goes quiet, the agent checks in, then closes the call politely instead of dropping it.
- Card digits stay out of logsA card number that appears in logs and transcripts as a valid 13 to 19 digit number is masked.
- Capacity you can plan forEach agent takes two calls at once by default. We raise it when your volume needs it.
Questions.
Do we need every capability?
No. Most teams start with one call type and the two or three capabilities it needs. Adding the next one later does not mean replacing the agent.
Can the agent transfer a caller to a person?
Yes, in your project. The agent rings the right person on your team (up to three destinations, each with its own "when to use it"), and once a person answers, connects the caller first, then introduces the caller and their request in one sentence both of them hear. It can also hand the case to your team with a written summary by email, or move the caller to a different workflow inside the same call. How transfer works.
Can it text callers?
Yes, in a project where texts are turned on: only to the caller's own number, after they agree, with your business name and an opt-out line added. Texts work on agents with a US number, once the carrier registration is approved. The agent can also send the caller a WhatsApp message, or email your own team.
How many calls can one agent take at once?
Two simultaneous calls by default. We raise the limit on request once we know your peak volume.
Each capability, in detail. How it works and where it stops.
Tool actions
Live lookups and approved actions in your own APIs during the call.
Tool actions →Outbound calling
Calls your systems place through the Voice API, with the result returned.
Outbound calling →Call intelligence
An outcome, a summary and any message the caller left, after every call.
Call intelligence →Handoff to your team
How a case that needs a person reaches your team, with the details.
Handoff →Background tasks
Longer work handed to your own OpenClaw agent on a private line.
Background tasks →Call audio
What the gateway does with call audio, and what affects clarity.
Call audio →Which call should your agent take first?
Tell us which requests your team handles most often. We'll suggest a practical place to start.