Noodle Seed - Your product in AI and AI in your product

@asadatnoodle nice launch!

I'm wondering how minified react components render across different mcp clients.
if the client can't render the ui elements,does it fall back to plain text schemas ?

@asadatnoodle avoiding all the framework work around MCP sounds like a big time saver for team,
Congrats on the team!

@tanjum yes and there are a few ways of doing this actually. We have baked in the ability to add a knowledge base document to your assistant which instructs and defines guard rails for the AI assistant on business rules, FAQs or any other relevant information. Moreover, you can also define an agent guide which is a skill that you can deploy alongside your MCP app and whenever you connect your MCP URL to any agent, it will receive that skill as well which helps it make better decisions around the tool usage and improve the overall conversational user experience.

@tanjum There are a number of ways the AI can be guided to represent your business or your brand the way you would like. For example, skills, knowledge bases, instructions, etc.
The most deterministic thing under your control would be small, minified user interface elements that can be sent over the wire to the user. Those elements are 100% deterministic and can be set up at the time the AI agent or the application being served through the AI agent is set up.

Even if the user is on Claude or ChatGPT, whenever one piece of user interface is sent to the user, it will always show exactly what you want: the same colors, the same font, and the same brand identity as you wish.

Would love to hear what kind of use cases you have in mind.

@monir_ I second that. There is a paradigm shift around user behavior as far as 'search interfaces' go. Agents are no longer question and answer bots anymore, they can take action and its only time when people will have to cater for this on their website, in their processes and workflows or overall operations.
We at Noodle Seed really believe in this and want to empower people to prepare for this wave by giving them as much technical flexibility as possible and enabling agent first experiences.

@monir_ An interesting point: we believe the conversation does not only evolve into the new search interface, but rather becomes a two-way, transactional interface, not a one-way interface or a one-way information interface, the way search used to be. People can not only ask for information but also complete entire transactions right inside the conversation, whether it's booking appointments, purchasing tickets, opening support cases, or anything else.

Would love to hear what use case you have in mind?

@1mirul Absolutely, it can!

We built the entire stack around the ability to showcase React-based components to the user, so we could not only show the products but also complete entire workflows and transactions right inside the AI conversation. Whether that conversation happens on your website, inside your web application, inside your mobile application, or inside someone's ChatGPT or Claude, the same capability can be delivered across all channels.

@shushantlakhyani Thank you Shushant, Join and start building with us !!!

@harini_mukesh Usually, customers start with an outcome, not a specific workflow: “Let me check inventory and place an order from ChatGPT.”

The workflows that translate best are frequent, well-bounded tasks where the intent is clear but the inputs vary. Search, lookups, updates, approvals, and simple multi-step actions work well conversationally. Dense visual work or high-risk actions still benefit from the existing product UI, with AI helping users get there faster rather than replacing it.

@harini_mukesh It’s usually a mix. Some customers arrive with a specific workflow in mind, but often we help identify the best place to start.

The strongest conversational workflows tend to be goal-based, easy to describe in plain language.
Some workflows naturally work better conversationally- especially tasks where users know the outcome they want but not necessarily the steps to get them there.

@angelaaa 
Here's my take on chatbots versus AI assistants: https://youtu.be/jWIkRheNxqU
The video will be live in a few hours.

A summary is that a chatbot answers questions, whereas an AI assistant helps your business grow and keeps your customers happy.

@angelaaa AI can serve a lot more functions beyond a mere chatbot widget. For it to be functional in a way thats valuable to you, its imperative that the AI interface lets customers come and interact with the business inside the conversation where you control the experience and make tools that help the AI convert that customer. This obviously goes beyond a traditional turn based conversation flow and allows the AI to create value for your business.

@protsenkoalexandra Claude already does organic discovery and suggests connectors from the indexed connector library and chatgpt is also in the works of achieving that. I believe they have some organic discovery already in place and it will recommend users relevant plugins based on intent.

With that being said, we believe users shouldn't be gate kept by OpenAI or Anthropic and should be able to get value on day one. We therefore also let users to setup an embeded assistant on their websites which is an out of the box offering from Noodle Seed that comes as part of your mcp plugin with minimal adjustments and a simple plug and play html script approach. If you however want much more control and want to design completely unique experiences, the assistant package lets you do that as well which still uses the same plumbing as the embeded assistant.

So while the big guys figure out discovery, your AI presence is already live and ready to serve your customers :)

@protsenkoalexandra For software companies in particular, or SaaS companies, they mostly drive their own customers to it. In fact, the customers almost demand an MCP interface, a Claude connector, a ChatGPT plugin, or something like that.
The way we have built this capability is that if you build it once on Noodle Seed, you get the copilot or embedded assistant inside the software almost for free anyway.