The useful question about ToqanClaw is not whether it can generate a dashboard from a prompt. Plenty of tools can now produce something that looks like software.
The useful question is whether Prosus can make AI-generated tools safe, connected, and boring enough for millions of restaurants, merchants, and small operators to use in real work.
That is the interesting part of this launch.
What Prosus announced
Prosus describes ToqanClaw as a platform that lets users create apps, dashboards, and automations by explaining what they need in normal language. It is built in-house, integrated with Prosus’ Toqan AI platform, and presented as an OpenClaw-inspired tool for business users who do not have engineering teams.
The pitch is ambitious. Prosus says it can bring the capability to more than 5 million restaurants, merchants, and entrepreneurs across its ecosystem. It also says the system routes work across 20-plus models, keeps data under customer control, and does not use that data to train third-party models.
That last sentence is doing a lot of work. In Europe especially, the difference between “fun AI demo” and “business infrastructure” is often governance: where the data goes, who can see it, who can approve actions, and what happens when the agent is wrong.
Why this is more than another no-code pitch
No-code has been around for years. The problem was never only interface complexity. It was context.
A restaurant owner does not want to become a workflow designer. A delivery operator does not want to map every edge case in a reporting tool. A small merchant does not want to understand model routing, permissions, and database joins.
They want to say: show me which delivery zones are wasting driver time, warn me when inventory is likely to break, summarize refunds by cause, or build a WhatsApp assistant for staff questions.
ToqanClaw is trying to make that intent-to-tool loop feel direct.
| Prosus claim | Why it matters |
|---|---|
| 5 million-plus merchants and restaurants | Distribution is the advantage, not just the model. |
| 60,000 internal agents and 10,000 applications | Prosus is claiming internal proof before partner rollout. |
| 20-plus model routing | Could control cost and match tasks to model strengths. |
| Secure Toqan environment | The product has to earn trust with business data. |
| Large Commerce Model | Prosus wants commerce context, not generic chat, to be the moat. |
The case studies are promising, but still company-reported
Prosus gives three examples: Lebkov & Sons cutting financial reporting from weeks to 30 minutes, Burger & Frites using a delivery analytics agent to raise deliveries and reduce overtime, and Poke Perfect reducing routine staff questions through a WhatsApp operations assistant.
Those are exactly the right use cases. They are not glamorous. They are operational. They sound like problems that real small and medium businesses actually have.
Still, they should be treated as launch claims until there is independent validation. AI vendors love outcome numbers, and business operations are messy. Revenue growth, overtime reduction, and reporting speed can all be influenced by many changes at once.
The promising signal is the category of work, not the exact percentages.
The OpenClaw comparison is useful, but incomplete
OpenClaw-style tools are popular because they shift AI from “answer this” to “do this.” They connect chat, apps, files, APIs, and actions. That is powerful, and also dangerous if permissions are sloppy.
Prosus’ argument is that ToqanClaw brings that pattern into a managed commerce environment. The advantage could be privacy, integrations, and domain context. The disadvantage could be lock-in, limited transparency, and the usual enterprise problem where “safe” also means slower to improve.
The winning version of this product needs a few things from day one:
| Requirement | Why it matters |
|---|---|
| Human approval for risky actions | A restaurant cannot have an agent changing prices or staffing blindly. |
| Clear audit trails | Owners need to know what changed, when, and why. |
| Permission tiers | A cashier, manager, and owner should not have the same agent powers. |
| Easy rollback | Generated tools will make mistakes. Recovery must be simple. |
| Plain explanations | Small businesses should not need an AI admin to understand the system. |
My take
I like the direction. A lot.
AI agents become more useful when they stop living as generic chat windows and start sitting inside real business loops. Prosus has a credible shot at that because it has distribution, transaction context, and direct access to operational pain across food delivery, classifieds, payments, travel, and local commerce.
The product could also become annoying very quickly if it turns into a black box that generates fragile workflows nobody understands. “Build any business tool from a single conversation” is a great headline, but the grown-up version is more modest: build a useful first version, connect it to the right data, test it, approve the risky parts, and keep improving it.
That is less magical. It is also how this becomes valuable.
Bottom line
ToqanClaw is worth watching because Prosus is not just selling AI as a productivity toy. It is trying to put agentic tooling in front of merchants who usually do not get enterprise software.
If the governance is real, ToqanClaw could be one of the more practical AI launches of the year. If the controls are weak, it will be another impressive demo that businesses are right to fear.
The difference will be auditability, permissions, and whether these generated tools keep working after the launch-day excitement fades.
Support the site at buymeacoffee.com/gearpulse.site.