WhatsApp Knowledge Agent
Answers customers on WhatsApp from your documents — and only the ones you published
- 9hard rules
- Things it must never do, enforced in every run
- 6handover rules
- Conditions where it stops and gives the work to a person
- 5output fields
- A structured result, not a paragraph you have to re-read
- ≤4steps per run
- Bounded work: at most 3 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Customers ask the same forty questions on WhatsApp, and answering them is somebody's whole morning. Most chatbots answer from the model's imagination, which is how a customer ends up quoting a policy you do not have.
This one answers from your documents. Publish a price list, a returns policy, opening hours or a product sheet to your WhatsApp knowledge base, and the agent replies from what is written there — in the customer's own language, in the short form WhatsApp is actually read in.
What it will not do is guess. If your published documents do not answer the question, it says so and offers a person, rather than producing something plausible. And it can only ever see what you published to WhatsApp specifically: internal documents in the same account stay invisible to it, because the restriction is enforced where the agent cannot reach it, not by asking the agent nicely.
Benefits
The repetitive half of the inbox, gone
Opening hours, stock, returns, directions — answered instantly, at midnight included.
Correct because it is quoting you
Update the document and every future answer changes with it.
Your team gets the real questions
They see the threads a person is actually needed for.
What it does
Answers from your documents
Replies come from the material you published, not from what the model believes about your business.
Only what you published to WhatsApp
Internal documents in the same account are unreachable. Publishing is per channel and deliberate, and enforced outside the agent.
Says when it does not know
An unanswerable question gets a handover, not an invention. This is the behaviour that makes it safe to leave running.
Written for WhatsApp
Short replies in the customer's language. Nobody reads a six-paragraph message on a phone.
Hands over cleanly
Asks for a person, mentions a complaint, or gets stuck twice — the thread goes to your team and the agent stops.
Every reply traceable
Each answer records the document behind it, so a wrong answer leads to the wrong document.
What this changes
What this changes about your week
- It takes over
- Answering the same forty customer questions on WhatsApp, all day, every day.
- Done today by
- Whoever has the phone — often the owner.
- How often
- Continuously, including at midnight.
- Typical manual effort
- The repetitive half of the inbox is somebody's whole morning.
Why Customer support in particular. It answers only from documents you published to WhatsApp, so a customer never receives a policy you do not have — and internal material stays invisible to it.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
WhatsApp Knowledge Agent has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-sonnet-5 with automatic fallback
- 3Toolssearch_knowledge
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output.
Input
{
"channel": "whatsapp",
"message": "مرحبا، هل عندكم توصيل للرياض وكم يستغرق؟"
}Output
{
"reply": "أهلاً بك 🙂 نعم، نوصّل إلى الرياض، والتوصيل يستغرق من ٢ إلى ٤ أيام عمل. تحب أساعدك بشي ثاني؟",
"escalate": false,
"disclaimer": "Answered only from your published documents, with the passage quoted. Anything they do not cover was handed to a person rather than answered, and nothing in a customer's message was treated as an instruction.",
"answeredFrom": [
"Shipping & Delivery Policy 2026"
]
}Requirements
Integrations
No integrations required.
Before you start
- A connected WhatsApp Business number
- A knowledge base published to WhatsApp, holding the documents customers ask about
- Requires the Starter plan or higher
Use cases
Retail and showrooms
Hours, location, stock and returns, answered from your own published sheet.
Clinics and services
What you offer, what it costs, what to bring — without booking or advising.
Property enquiries
Answer from the published listing sheet, then hand a serious enquiry to an agent.
After-sales support
Warranty terms and setup steps, quoted from the documentation you already wrote.
Pricing
Starter
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Can it see our internal documents?
No. It reads only knowledge bases you published to WhatsApp. Everything else in your account is invisible to it, and that restriction is enforced in the retrieval layer — not by an instruction the agent could be talked out of.
What happens when it does not know?
It says so and offers a person. It is instructed never to answer a question your published documents do not cover, because a confident wrong answer on WhatsApp is worse than no answer.
Which languages?
It replies in whatever language the customer wrote in, from documents in whatever language you published.
Can it take orders or payments?
No. It answers questions and hands over. It cannot place an order, take a payment, make a booking, or promise anything on your behalf.
What if a customer tries to trick it?
Customer messages are treated as questions, never as instructions. A message telling it to ignore its rules or reveal internal information is answered as what it is — a question it will not answer — and the retrieval limit holds regardless, because the agent has no access to widen.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent