VATES

VATES puts your company’s knowledge base inside your own site or app, so a visitor asks a question in their language and gets the answer in that same language — even when the knowledge itself is written in a completely different one. It is built by STANDOUT Inc., based in Okayama, Japan.

What makes it worth a look from an Arabic-speaking angle specifically: the product page in Arabic is written in genuinely correct Arabic, not machine translation. That is a practical signal, because a tool that will not bother writing proper Arabic on its own marketing page rarely handles Arabic well inside the product.

The core idea: knowledge in one language, answers in the asker’s

Suppose you run a technical company in Cairo and your product documentation is in English because that is what your team works in. A visitor types a question in Arabic. Conventional systems leave you two options: translate all your documentation into Arabic and carry the burden of keeping two versions in sync, or let the visitor receive an English answer they cannot read.

VATES proposes a third path. You keep your knowledge in one language — whichever you like — and the system answers each asker in theirs. A Japanese company writes in Japanese; its Arabic-speaking visitor gets Arabic, without the company doubling its work.

This is not unprecedented; plenty of systems do something similar today. The distinguishing detail is that answering in the asker’s language is not a feature bolted on later here — it is the central design point the product was built around.

Three roles: bard, druid and vates

The tool divides itself into three components named from Celtic tradition. The naming is more thematic than technical, but the roles are clear:

• bard — the receiver. It takes whatever you give it (natural language, logical symbols, images, PDFs) and converts it into an internal format the company calls ES, short for EchoScript.
• druid — the management interface. This is where you review, edit and full-text search the knowledge base, and read usage logs.
• vates — the responder. It consults the knowledge base and composes the reply in the asker’s language.

In practice you deal with druid while setting things up, and your visitors deal with vates without ever knowing it is there.

Why “no vector retrieval” actually matters to you

The company states plainly that the system runs with no vector retrieval, no embedding pipeline and no re-ranking. That is technical language, but the consequences for you are concrete.

In the common embedding-based systems, your documents are converted into numerical representations stored in a vector database. That brings two irritations. First, editing a document requires re-indexing, which can take time before the change has any effect. Second, what the system retrieves is not text you can read — it is numbers that mean nothing to you, so when the system gives a strange answer, tracing where it came from is difficult.

The VATES approach evaluates knowledge units textually at question time. Two direct results: an edit takes effect immediately with no re-indexing wait, and what gets retrieved is human-readable text you can inspect with your own eyes to understand why the system answered as it did.

Be realistic about the trade-off, though. This approach suits curated knowledge bases of moderate size — FAQs, product documentation, company policies. Anyone with hundreds of thousands of unstructured documents will likely find vector-based systems a better fit. The tool is making a clear trade, not claiming superiority in every case.

Three ways to use it

• A ready-made widget you drop into your site; the company notes it is CSS-driven, meaning light on the page.
• A standalone app if you want a self-contained interface rather than an embed.
• An API for developers, together with a remote MCP server using OAuth 2.1 — which in practice means AI agents (those built on the MCP protocol) can connect to your knowledge base directly and with proper authentication.

One point worth noting for anyone who runs their business from a phone: the company states the entire console — setup, billing, security and knowledge-base editing — is built to work on mobile, rather than being a shrunken version of a desktop interface.

Pricing

The model is deposit-based rather than subscription-based: you top up a balance, and usage is deducted from it. The company states explicitly that there is no monthly subscription, no minimum spend and no cancellation fee.

A company representative stated the minimum deposit is five US dollars. We could not confirm that figure from the pricing page directly, because it loads via JavaScript and the number does not appear in the static page source — so treat it as the company’s figure rather than one we verified ourselves, and check it at sign-up.

Either way, the absence of a monthly subscription makes trying it low-risk: you can put in a small amount and test the tool against your real questions before committing to anything.

Who is behind it

The developer is STANDOUT Inc. (株式会社STANDOUT), based in Okayama, Japan, with its full address published in the structured data on its own site. That is a higher level of transparency than many tools listed in directories, where who runs them and from which country is often unknown.

The company states the tool is available in the ChatGPT Plugin Directory, and that the infrastructure includes hash-chained tamper-evident audit logs, passkeys, two-factor authentication, IP restrictions and per-tenant rate limiting.

Who it suits, and who it does not

It suits you if: you run a site or app serving a multilingual audience and have curated knowledge (FAQs, documentation, policies) you want answered automatically; or you want first-line customer service in Arabic without translating all your documentation; or you are a developer building an agent that needs knowledge-base access over MCP.

It does not suit you if: you are looking for a general chatbot that answers anything — this one answers from your knowledge, not the world’s; or you have a large, unstructured archive you want to search semantically, which this textual approach is not designed for; or you need a product with an established reputation and a broad user base, since this is a relatively small company and new to the global market.

A note on transparency

This tool was submitted to us by its developer through our tool-submission page. We reviewed their site and verified what could be verified — the existence of pages in twenty-three languages, the technical claims published on the site, and the company details — and we have flagged above what we could not confirm. We received no payment for this listing, and we have not yet run an extended hands-on trial; treat the above as a review of what is stated, not a performance endorsement.