A studio lives or dies on proof. The old Webanion site was a set of hand-built pages: every new project meant editing several of them, the writing behind the work never made it onto the site, and a buyer reading in Bengali, Spanish, Hindi or Chinese got English or nothing. Worse, buyers had started asking AI assistants for a shortlist, and those assistants were guessing from scraped pages that were already out of date.
The answer was to stop treating the portfolio as pages and build it as a product: one source of truth for every case study, served to people in their language and to assistants in a form they can trust.
### Every case study is a record, not a pageProjects, services, companies, clients, testimonials and articles are structured records in a headless CMS, linked to each other. Change a record once and every page that mentions it updates: the project page, the company page, the service page, the search results.
For a buyer that means the proof is always current and always connected, from a service they need to the projects that show it and the people who vouch for it. For the founder it means a new project is published once rather than copied into an afternoon of pages.
### Five languages from the same recordEnglish leads, and every record also carries its own Bengali, Spanish, Hindi and Chinese version, rendered from the same structure at every screen size. A buyer in Dhaka, Madrid, Delhi or Shanghai reads the work in their own language with the same depth as the English, instead of a translated homepage with English everywhere behind it.
### A story is the project page, cover to closeCase studies that matter get a full story: a cover, then sections stacked on one canvas, each with its own layout and animated artwork, all written and arranged in the same CMS. The story is the project page itself, so a buyer who lands from search reads the problem, the solution and the result without leaving it. Stories are data too, so they translate like everything else and change without a developer.
### An MCP server so assistants read the sourceAlongside the website runs an MCP server that exposes the same content to AI assistants as typed tools: search the portfolio, read a project, list services, check a testimonial. An assistant asked about Webanion reads the record rather than guessing from a scrape, and a buyer who wants more can request a key and send a message through the assistant itself.
That makes the portfolio a lead channel in a place most studios are not yet present, and every answer an assistant gives is one the studio wrote.
### Private by constructionA portfolio holds things that must never leak: client contacts, confidential companies, drafts not yet approved. So the public surface is an allowlist decided field by field in one place, enforced in CI, and anything not on it cannot be returned by the website or the MCP server. Drafts stay invisible until they are published, and the owner's admin surface sits behind three independent checks.
Clients under NDA can be shown as proof without being named, which is often the difference between having a case study and not.
### Nothing goes live without a readEvery change lands as a draft first, whether it was typed in the admin or drafted by an AI assistant through the owner's own MCP server. The founder reads it in a private preview of the real page, hidden from search engines, and only then publishes. A one-person studio keeps its portfolio as current as an agency's without shipping a word nobody read.
### Proof that compoundsEvery project added to Webanion lands on the right pages, in five languages, searchable on the site and answerable by assistants, the same day. For buyers it is the fastest way to see relevant work they can trust; for the studio it is a portfolio that grows in value with every record instead of going stale.
The platform runs on the same self-hosted Kubernetes cluster as the client work it describes, told in the zero-downtime migration story, and the assistants that help run it are described in the AI engineering story.