How to integrate “The #EverythingAI Agent” with anything?

What if you could have an AI agent which can help you with anything and everything? Sort out your holidays, fill out paperwork, find the best home insurance offer, do the grocery shopping, buy flowers for mother’s day, schedule your vacuum cleaning robot, …

To have an everything AI agent, the agent needs to know thousands of processes, know millions of facts and be integrated into thousands of systems. No single company would ever be able to build such an agent. So what if we have thousands of companies join forces to build “The Everything AI Agent”? Some offer grocery delivery. Others flowers. Insurance comparison. Travel. Payments. All have APIs. So what is needed is a way for each of these companies to integrate their solution and allow The Everything AI Agent to interact with it.

Luckily there is an open source standard called Model Context Protocol or MCP which allows AI agents to dynamically discover new tools, data sources,... MCP enables The Everything AI Agent to call a weather API to get a weather forecast, Amadeus APIs to book flights, Uber to go to the airport. AirBnB to book accommodation. OpenTable for a restaurant. TripAdvisor for a tour guide.

So if every company with an API would enable theirs to be integrated via MCP into any AI agent, we would be a huge step closer to The #EverythingAI Agent. But how would you run thousands of API integrations written by thousands of companies? This is where a new open source project called MCP-Ectors has come up with a unique solution. Why not package all of this code into secure containers so that just like a container ship that carries thousands of containers, The #EverythingAI Agent can access thousands of integrations in a standardised way.

Many IT people know of Docker but these containers can easily be hundreds of Megabytes, if not Gigabytes. We need something that is more similar to a mobile app. The solution is called WebAssembly or WASM for short. WASM containers are small [from Kilobytes to tens of Megabytes at most]. Through a standard called WebAssembly System Interface or WASI, these “code containers” can expose standardised interfaces which allow external systems to talk to them in a unified way. You can package badly written code, even hostile code like a virus inside a Wasm. The only way it will be able to talk to the outside world is via WASI. So unless you give it access to a directory or a URL, it will not be able to interact with anything.

MCP-Ectors has just open sourced the WASIx-MCP interface which allows any company to package the code to call their API and offer it via an MCP to any AI agent. MCP-Ectors afterwards allows you to drop in your Wasm into a directory and enable the AI agent to discover it and interact with it. Soon companies will be able to upload their Wasms to an AI Store and any AI agent to get it from there. AI agents will be able to dynamically extend their capabilities based on what the human or other AI agents want it to do.

If you are interested in helping to shape The #EverythingAI Agent revolution you can:

• Subscribe and be kept up to date as well as give MCP-Ectors a star.
• If you want to drive where The #EverythingAI Agent is going, why don’t you become a sponsor.
• If you know how to programme, help extending MCP-Ectors or start creating your first MCP Wasm. Here is an example.