With every new technology platform, the concept of an application is shifting. Consider the difference between compiled apps during the PC era, online applications during the web and app stores during mobile. Now it happens again with AI.
Before AI is in the impact on applications, it is worth mentioning that we still have downloadable desktop applications, web applications, mobile app shops and everything in between. Technology platform does not shift the past and they do not happen from one day to the next. So ai-driven changes, while they happen quickly, will happen for a long time.
The basic components of an application have also remained consistent for a long time. An application at the highest level is simply code and a database. The database stores the information that manipulates an application and with the current code you can manipulate it through input and output checks (user interface, auth, etc.).
As AI coding agents have become more capable, they have increasingly been able to process more of the current code aspect of an application. They can not only generate code, they can view, repair and maintain it. It is therefore not difficult to see how AI agents can be a self-sufficient loop.
As AI coding agents take on the current code aspect of an application, as they increasingly have to make, update and work with databases. Today’s databases are made for people to use, not for agents. So we have built a database system for AI applications called Agentdb Designed for agents, not for people.
Agentdb With this, agents can manifest new databases by only referring to a unique ID. Instead of filling in a series of forms – as people do when making a database. It also offers agents templates they leave Start using Databases immediately and consistent about use cases. These templates are dynamic, so because agents learn new or better ways to use a database, information is passed on to all the following use of agent.
With these two changes, the concept of an application is already shifting. But what if the idea of ​​needing “current code” also changes? By giving an agentDB database and the template system with an external model context protocol (MCP) server to the head: only you are an URL plus an AI model to have an app.
All you need is a URL Plus an AI model to have an app.
In this video I demonstrate the uploading of a CSV file from a credit card statement to AgentDB. The system makes a database and template, summarize both with an external MCP server -URL that you can add to any AI application that supports external MCP such as Claude, Cursor, Augment code, etc. The end result is an immediate chat app.
Through natural language instructions you can read and write data immediately and consistently for each variant of user interface you want. Most credit card websites are painfully restrictive, but now I can make the specific visualizations, categories, questions and functions that I want. Don’t wait until the credit card site implement new code.
You also do not need a CSV file to make an app. Just tell an AI model that is connected to agentdb what you want. It can use AgentDB to create, fill a database and to ensure that everything you add contains the correct information. Following the date, location and costs of concert tickets? AgentDB will enforce all the information that is there and if you add a new bit of data to keep track of, this can update all your records (see video below).
You can try to make your own chat app from a database or CSV file on The Demo page On AgentDB to get an idea of ​​it. There are definitely some rough edges, especially when you try to add an external MCP server to a few AI applications (in fact this whole step must disappear), but it is still pretty compelling.
As I said in the beginning, we do not fully know how the AI ​​platform shift will transform the applications. However, it is clear that there are major changes.
#Lukew #Reconsideration #applications