Rendreal connects an AI conversation to a physical print order. A customer describes what they need, creates or uploads artwork, picks print specifications, and finishes with a checkout link, all in one conversation. Rendreal builds on experience from Mixam's web-to-print business, and partnered with Vstorm to build the system. Vstorm, an official Pydantic implementation partner, orchestrated every agent with Pydantic AI, and the tools reach outside assistants through an MCP server.
The challenge: An assistant can design a leaflet, but it cannot print one
Ordering print still means leaving the conversation. A customer opens a printer's website, works through a product configurator, and then either designs artwork in a browser tool, pays a designer, or uploads a file made somewhere else. Someone then checks that file against the product: bleed, resolution, and whether the text fits the space. When the file fails, the order waits.
General-purpose AI assistants close part of that gap. Ask one for a leaflet and you get something that looks good on screen. It is not a print file, it has no price, and there is nowhere to send it.
Mixam had already shown that an agent can guide a customer through complex print specifications on the printer's own website. Rendreal asks the next question: what if the customer never visits the printer's website at all?
| Mixam | Rendreal | |
|---|---|---|
| Access | A website with an agent that advises on and takes the order | Compatible AI assistants, web chat, or WhatsApp |
| Order | Prints from existing content | Prints the AI can design from the customer's assets, as well as existing content |
| Fulfillment | Supplier network | Integrated print suppliers |
The solution: One set of tools, three ways in
Customers reach Rendreal through three channels:
- Web chat, the main channel, with a conversational interface plus interactive product and design controls.
- WhatsApp, where a customer can request a design from a bot without opening a website.
- External AI assistants, which connect to Rendreal's MCP server and call its tools directly.
Each channel has its own entry point and conversation handling. All three share the same business services for products, designs, and orders, so a configuration means the same thing wherever the customer starts.
The flow is the same in every channel. The assistant finds a product. If the customer needs artwork, Rendreal's design service generates or revises it for review. The assistant requests a quote, configures the product, creates the order, and returns a checkout link. The customer pays on a secure checkout page, and printing and delivery run through Rendreal's supplier integrations.
Product options come from the integrated catalog. Rendreal checks every configuration against those options and tells the customer when a request is something the product cannot support. Pydantic AI does the first part of that work: it validates tool arguments and structured outputs against typed schemas, and retries when a model returns something malformed. That check matters most at the last step, because the finished specification has to pass Rendreal's order validation before the order is taken in.
Specialist agents behind the MCP tools
The Model Context Protocol lets a compatible assistant discover and call external tools. Some of Rendreal's tools are simple lookups. Others hand the work to specialist AI agents, including one for product configuration and one for design.
Vstorm orchestrated the main agent and every specialist agent with Pydantic AI, so all of them share the same typed tool definitions and the same validation on the way out.
It helps to be precise about what this is. Rendreal gives assistants tool access to its own integrated services. It is not an open network of independent supplier agents that find each other and negotiate orders. That longer-term picture, where an assistant finds service providers and arranges more of the work on the user's behalf, is where Rendreal is heading, not what it claims today.
Print-ready designs from a request
Taking print orders through an assistant removes most of the friction. One obstacle usually remains: a print-ready design. Rendreal lets customers drop in their assets, describe the goal, and get a printable file for:
- Flyers, including folded leaflets
- Posters
- Business cards
- Postcards
- Bookmarks, including two-sided versions
Each product brings its own constraints, from tight space to keeping typography consistent across panels. The design service handles them through five capabilities:
| Capability | How it works |
|---|---|
| Generate from a prompt | The customer describes what they need, and the service proposes artwork. It may ask for more detail, and the customer can request revisions. |
| Use reference material | Copy or images can be supplied as direction, not to be reused as they are. |
| Edit the current design | The tool can move elements, reshape lettering, or update copy. It supports full revisions and targeted edits, and the customer reviews the result before ordering. |
| Switch visual styles | The customer can browse predefined styles for inspiration, or start from scratch. |
| Snapshot versions | Versions are saved along the way, so the overall direction is not lost. |
The result is that a technical, error-prone pre-print step becomes a tool the customer's agent can call, or that the customer can use directly.
What's next
The stack is ready for model changes. Because Pydantic AI is model-agnostic, moving to a newer model does not mean rewriting the agent logic. The evaluation set stays with the system, so the same test can run again when new models ship.
The results
Today, Rendreal offers a working path from a sentence to a paid print order across three channels and five product types, with every configuration checked against the catalog before a quote is issued.
Key takeaways
- Treat the printer as a tool, not a destination. Exposing ordering through an MCP server lets customers stay in the assistant they already use.
- Validate before you quote. Typed schemas in Pydantic AI catch malformed output inside the agent loop, and the catalog check stops impossible orders before they cost anyone time.
- Split the entry points, share the business layer. Web chat, WhatsApp, and MCP each handle conversations their own way, but products, designs, and orders live in one place.
- Say exactly what your agents do. Tool access to integrated services is useful on its own. It does not need to be described as an autonomous agent network.
Building agents that other assistants can call? See how Pydantic AI works with MCP or read the full case study on vstorm.co.
Frequently asked questions
What is Rendreal?
Rendreal turns an AI conversation into a physical print order. Customers describe what they need, get print-ready artwork and a quote, and finish with a checkout link, in web chat, WhatsApp, or a compatible AI assistant.
How does an AI assistant connect to Rendreal?
Through Rendreal's MCP server. Any assistant that supports the Model Context Protocol can discover Rendreal's tools and call them to find products, generate artwork, get quotes, and create orders.
Is Rendreal an agent-to-agent network?
No. Assistants call Rendreal's tools, and some of those tools delegate to Rendreal's own specialist agents. Independent supplier agents do not discover or negotiate with each other.
Does Rendreal use Pydantic AI?
Yes. Vstorm orchestrated every Rendreal agent with Pydantic AI, which validates tool arguments and structured outputs against typed schemas and retries on malformed responses. Because the framework is model-agnostic, Rendreal can change models without rewriting its agents.
