Multi-tenant WhatsApp for agencies
One dashboard. Every client's WhatsApp. Zero data bleed.
Run 5 to 50 client brands from a single control plane — each workspace isolated at the query layer, each with its own inbox, flows, AI, and campaigns. Built for the agency, not retrofitted for it.
We don't have a logo wall yet. We have proof you can inspect.
real business runs its production WhatsApp number on Rasilo today — Fixlr, our smart-home installation marketplace, with AI-first 24/7 support.
first-party brands run on Rasilo workspaces — Fixlr, Lumive, and Ribat. Ours, labeled as such, not dressed up as customer stories.
of the screenshots on this page are the real product. The conversations use demo data; the flows and AI are the ones we run.
A live isolation demo
Watch Workspace A's API key ask for Workspace B's data and get a hard denial — the exact response is shown below.
A self-guided tour
Walk every capability below — no sales call, no signup, no invented dashboards.
A public API you can read
Contacts, template sends, template management, signed webhooks — documented with an OpenAPI spec, shown honestly below.
The isolation promise
"Zero data bleed" isn't a slogan. Here's the mechanism.
Every request carries an API key scoped to exactly one workspace, enforced at the query layer — not in the UI, not by convention. Ask for another workspace's record and it simply isn't in scope: a hard 404, the same as if it never existed. Not a filtered list. Not a partial leak.
And the secrets stay sealed. Each workspace's provider credentials — LLM keys, channel and email secrets — are encrypted at rest, per workspace, so one tenant's curiosity never becomes another's exposure.
GET /v1/contacts/ctc_A1 · Bearer ck_workspaceA
200 OK
GET /v1/contacts/ctc_A1 · Bearer ck_bad
401 Unauthorized
GET /v1/contacts/ctc_B7 · Bearer ck_workspaceA
404 Not Found
401 means "we don't know who you are." 404 means "that isn't yours, and we won't even confirm it exists." Cross-tenant reads always resolve to 404 — never a 403 that would leak the record's existence.
Want to poke at the isolation yourself? A guided setup includes a sandbox workspace and a key of your own.
Book a guided setupOne control plane
The whole WhatsApp lifecycle, in one place.
Every image below is the actual product — no illustration, no invented UI. Conversations use demo data; the flows and AI are the ones running on our own brands.
Answer as a team, hand off honestly
A WhatsApp-first shared inbox. The moment an agent replies, the AI pauses automatically — a graceful handoff, not a toggle someone forgets. The 24-hour window state is always visible, so nobody drafts a message that can't be delivered.
In the shot: bot answers, then a human takes over — AI paused, window open, one thread.
Build flows you can actually see
15+ node types, an AI drafter, a simulator, and versioned publishing. The diagram view below is our real Fixlr assistant — a bilingual router with menus, AI answers, and human-escalation paths — 66 nodes, running in production.
In the shot: the live Fixlr — Assistant flow, escalation edges in red.
AI that answers from your content
Feed a knowledge base — text, PDFs, or taught Q&As — and the agent answers from it, with citations and per-agent escalation rules: not-in-KB, low confidence, "get me a human," repeated failures. Add a vision model and it reads customer photos too.
In the shot: our production agent configuration.
Broadcasts that respect the rules
Template campaigns with audience filters and a live recipient count, per-recipient variables, scheduling, throttling, and delivery tracking — plus parallel email campaigns with SES senders and suppression lists.
In the shot: the campaigns view of one of our own workspaces.
A CRM that grows its own list
Bulk CSV import with mapping and a consent gate, custom fields, tags, lifecycle stages and opt-in status — plus QR codes and click-to-chat widgets that tag and attribute every contact they bring in.
In the shot: the contacts table (demo data).
Built for teams — and for Arabic
Role-based access with per-module overrides, so an agent sees the inbox and nothing else. And the entire product is bilingual English + Arabic with full right-to-left mirroring — not a machine-translated afterthought.
In the shot: the same dashboard, in Arabic, fully mirrored.
Works with
How it works
From zero to a live client workspace in three steps.
Spin up an isolated workspace
Each client gets its own tenant — its own inbox, contacts, flows, AI, and encrypted credentials. Nothing crosses between them.
Connect your channel, pick your model
We set up your WhatsApp connection with you (guided today) and you choose your LLM per workspace — so you never stare at an empty screen.
Build, automate, and hand off
Draft a flow with AI, ground your agent in a knowledge base, launch a broadcast — and let the AI answer until a human needs to step in.
Three ways agencies run it
Same control plane, different clients.
A customer scans a table QR → lands in WhatsApp, tagged by branch → a flow takes the order details and a human confirms.
Growth widgets attribute every contact to the branch that brought them in.
A lead clicks a listing ad → the AI answers project questions from the brochure KB → "talk to an agent" hands off to the sales inbox.
Escalation rules make sure a human gets every serious buyer.
Order updates go out as template campaigns → replies land in the shared inbox → a post-purchase flow collects a rating a week later.
Message-based pricing means a big contact list costs nothing extra.
Compare honestly
Where Rasilo actually differs.
Generic columns on purpose — we compare against categories, not named competitors we'd have to caricature.
| Capability | Rasilo | Typical BSP dashboard | DIY on raw APIs |
|---|---|---|---|
| True per-tenant isolation (query-layer, hard 404) | Yes | Varies — often UI-level | You build it |
| Bring your own LLM (Claude / OpenAI / Ollama / OpenRouter) | Yes | Rare — usually one vendor | You build it |
| Message-based pricing (no per-contact tax) | Yes | Often per-contact tiers | Yes |
| Full Arabic RTL product UI | Yes | Rare | You build it |
| Public API + HMAC-signed webhooks | Yes | Varies | Native |
| Inbox, flows, RAG AI, campaigns out of the box | Yes | Mostly | Months of work |
Candor, not badges
Where we are — honestly.
Rasilo is pre-GA. We'd rather you know exactly what's shipped and what's coming than discover it after signup.
Shipped and working today
- Multi-tenant workspaces, query-layer isolation, per-workspace encrypted secrets
- Inbox, flow builder, in-chat forms + Sheets, RAG AI, broadcast + email, CRM
- Pluggable LLM per workspace: Claude / OpenAI / Ollama / OpenRouter
- Public /v1 API: contacts, transactional template sends, template management, signed webhooks + OpenAPI spec
- RBAC with per-module overrides · bilingual EN + Arabic RTL
On the roadmap — labeled honestly
- Meta Cloud API / bring-your-own-number & channel-provider choice
- Self-serve connect-number wizard (today, setup is guided)
- Automated billing (Stripe), published tiers, white-label / reseller
- Per-message delivery-status webhooks (today: send results + system events)
- SSO / SAML, MFA, admin audit log · signed DPA, data-residency, SLA
For your developers
A real API, not a checkbox.
Your backend triggers a template send with one call — OTPs, order confirmations, lifecycle notifications. Rasilo owns the WhatsApp plumbing; you keep your own system of record.
- Idempotent sends — a retried request with the same dedupe key can never double-message a customer.
- Template CRUD over the API, mirrored to your WhatsApp provider, with change events broadcast to your webhook.
- HMAC-SHA256-signed webhooks with retries and dead-lettering — verify with ten lines of code.
- Scoped API keys per integration, an OpenAPI 3 spec, and an integration guide with real captured responses.
curl -X POST https://api.yourdomain.com/v1/messages/template \ -H "Authorization: Bearer ck_…" \ -d '{ "to": "+9665xxxxxxx", "templateId": "tpl_order_confirmed", "variables": { "order": "8241" }, "dedupeKey": "order-8241-confirmed" }'
200 { "messageId": "…", "status": "sent", "deduped": false }
Pricing
Message-based, not per-contact.
WhatsApp moved to per-message pricing in 2025; Rasilo aligns to it. You pay a platform fee plus pass-through message costs — no per-contact tax that punishes you for growing your list.
| Line item | Who charges it | Notes |
|---|---|---|
| Rasilo platform fee | Rasilo | Per workspace / tier · month-to-month or annual |
| WhatsApp message costs | Meta / your provider | Billed at provider rates — we don't mark them up |
| Email sending | Your SMTP / SES | You own the sender and its cost |
Committed, published tiers are part of our GA path. Until then we won't invent a number — a guided setup includes a transparent cost walkthrough for your exact client mix, and founding-cohort workspaces keep their setup pricing when tiers publish.
Objections, answered plainly
Frequently asked questions.
Is this compliant with WhatsApp's rules?
Can I migrate an existing WhatsApp number?
What happens to my data if I leave?
What does "guided setup" cost and involve?
How are secrets and credentials handled?
Who owns the AI model relationship?
How deep does the Arabic support go?
When does self-serve signup land?
Run every client's WhatsApp from one place — and see the isolation for yourself.
No logo wall, no invented metrics. A control plane built around multi-tenancy, pluggable models, and honest handoff — with a team that will set it up with you.
Or just write to support@rasilo.app — a person answers, not a sequence.
We onboard a limited number of guided setups each month · Pre-GA · Bilingual EN + AR / RTL · Your model, your data