Prostir

About Prostir

Turn what you know into an AI Agent your audience can access.

Most of what you know already sits in your files and in the answers you repeat every week. Turning that into something other people can use is where it gets complicated — publishing it, deciding who gets in, setting limits, taking payment.

Prostir takes care of that part. You bring the material and say how it should answer.

01Why we built it

Why we built Prostir

It began as a small internal tool. We needed an AI agent to take over part of our own finance workflow, and the brief was clear enough that we could describe it in one sentence and build most of it ourselves.

Describing what it should do turned out to be the easy part. Sharing it was not. Access rights, who sees what, updates that have to reach everyone at once — we ended up pulling in our own tech team for decisions that had nothing to do with the idea and everything to do with the machinery underneath it. It kept feeling harder than it should have been.

The distance between knowing exactly what you want and having it actually running was full of work none of us saw from the outside. So we removed that technical layer for ourselves, once and for all.

Then we asked whether it was just us. People in consulting, at a clinic, in marketing kept giving the same answer: the idea was never what stopped them. So we built Prostir to take that layer off other people too: one place to publish what you know, decide who gets in, set limits, and keep your material in order — so a good idea does not sit in drafts while the same questions keep coming back.

02What it is

What Prostir is, in plain words

You add what you already know — documents, notes, the answers you repeat every week. You describe how it should answer, in your own words. Then you publish, and it gets its own link that people can talk to, plus a connection that AI apps like ChatGPT or Claude can use when they support it. Putting it on your website or in a chat app is a separate step.

  1. 01
    Add what you knowDocuments, notes, guides — whatever you answer from today. Prostir reads the text inside them, so the clearer the file, the better the answers.
  2. 02
    Say how it should answerIn your own words. No code anywhere.
  3. 03
    Choose who gets in, and what it costsKeep it free, or charge for it and get paid into your own Stripe or Rozetka Pay account.
  4. 04
    Publish itYou get a link you can share, and it keeps working while you keep editing.

What you can switch on:

  • Sign-in, so you know who is asking
  • One-off or subscription payments through Stripe
  • One-off payments in UAH through Rozetka Pay
  • Usage limits, and a view of what people ask

What you can switch on depends on your own setup, whether you qualify to sell, and what we have finished building

03Who it's for

Who it is for

Prostir is built for people who have the expertise but not an engineering team.

  • Consultants & coaches

    Who answer the same questions every week.

  • Course creators

    Who want a tutor that teaches from their material.

  • Teams

    Who need a private helper that knows their docs and rules.

  • Anyone with a service to sell

    Who wants to charge for what they know without building the plumbing first.

Start by checking that your files, the way you want to share it, and how you want to get paid are supported.

04

Who builds it

Prostir is made by Managed Code, a creative and engineering team that builds products. We have spent years building the parts most products end up rebuilding from scratch: hosting, search over your own material, payments, and who is allowed in.

We are turning that work into pieces anyone can reuse. Some of them are further along than others, and we say so rather than pretending everything is finished.

If you want to see exactly what runs under the hood, the next section has it.

05Under the hood

For the technically curious

You do not need to read this to use Prostir. If you build software, here is what runs underneath.

Prostir is a .NET platform built by Managed Code. The creator workspace and operator console run on Blazor; the public pages are static Astro. Underneath, typed .NET services on Orleans handle the boundaries between agent, user, session, and operation, with Cosmos for persistence and EF Core for the query model. Aspire orchestrates it locally and in the cloud.

The AI layer uses Microsoft.Extensions.AI for chat clients, embeddings, tool calls, and telemetry. Three Managed Code packages do the heavy lifting: MCPGateway aggregates tools, prompts, and upstream MCP sources; MarkdownLd.Kb builds the Markdown-LD and JSON-LD knowledge artifacts behind search; Storage abstracts files, memory, and runtime artifacts.

Actions stay server-side. The official MCP C# SDK handles protocol, Jint runs sandboxed JavaScript tools, and Stateless drives workflow transitions and can-fire checks. Eligible seller flows use Stripe Connect for one-time payments or subscriptions; Rozetka Pay marketplace checkout is Ukrainian, UAH, and one-time only. The seller remains responsible for tax, fulfillment, refunds, disputes, and warranties. Prostir does not offer pay-per-use billing.

  • MCPGateway
  • MarkdownLd.Kb
  • Storage
  • Orleans
  • Cosmos DB
  • EF Core
  • Aspire
  • Microsoft.Extensions.AI
  • MCP C# SDK
  • Jint
  • Stateless
  • Stripe and Rozetka Pay integration

Where to start

Explore examples before you pick a build path.

Examples illustrate possible Agent shapes without claiming customer results. A case study belongs here only after the customer, implementation, and outcome evidence have been verified.