AI Agent2Agent (A2A), Model Context Protocol (MCP Servers) & Visual Programming Language

Author(s): Andrei Besleaga (Nicolae)

Originally published on Towards AI.

Sharing new experiments, as libraries SDK implementations of developments, from the world of AI advancements: Agent2Agent (A2A — Google) and Model Context Protocol (MCP — Anthropic) protocols, standards to be, which allow communication between AI agents, and connecting them with other sources via MCP servers, (which in this case refers also to some open EV charging protocols), which will further enable AI assistants (and any other programs where they are integrated), to interact with external systems sources (like EV charging infrastructure), for advanced usages.

My experiments (as examples, of which I could not find any implementations yet, on which anyone is welcomed to help, or use):

• a2a-php — A2A PHP SDK — PHP implementation of the A2A (Agent-to-Agent) protocol, library which tries to implement the protocol in PHP, and make it available to other php projects to create and use agent-to-agent flows, easily, via composer.
• ocm-sdk — OCM SDK & AI MCP Server — TypeScript implementation for Open Charge Map API.
• Demo for OCM MCP lib https://github.com/andreibesleaga/ocm-demo, at: https://ocm-demo.onrender.com/
• ocpi-sdk — OCPI SDK & AI MCP Server — TypeScript implementation for Open Charge Point Interface
• ochp-mcp — OCHP AI MCP Server — TypeScript implementation for Open Clearing House Protocol
• oscp-sdk — OSCP SDK & AI MCP Server — TypeScript implementation for Open Smart Charging Protocol
• The missing projects (not yet started, but that might be useful at some point, or not) : MCP PHP SDK libraries and other useful electric vehicle charging protocols (OCPP, OpenV2G, OpenADR, OVMS, etc.), and any other protocol/data/system source that could be connected in a MCP standardized way, to other AI assistants, agents, and systems.

Most of them are developed with help from AI or other automated tools (Stainless), based on schemas of the protocols already available, so we can also see the usefulness of standards as OpenAPI.

More than these, a Systems Software Architecture for a Collaborative Visual Programming Language with agentic AI and other hardware integrations, including systems for easy development and deployment, is described here : https://github.com/andreibesleaga/VisualGridDev.

Of course most of these above projects have no real worldwide application usefulness yet, and there are also other classical ways to implement anything as those, or combined ways (this project is another example), but these protocols and others are supposed to become standardized ways to glue some of the final AI things (agents) together, and probably more will come or change, as AI systems architectures are also evolving, and these example implementations are examples of how innovative aspects, protocols, and usages, could be taken in consideration in the systems of the future, which might embed both classical real-world apps with AI developments.

The positive outcome is that already having such a large base of projects, based on these new protocols, might not even require any new libs or programs developed, but with some simple clicks to auto-generate the most optimal A2A communication flow or MCP Server for any conceivable system, using the most performant scalable solution behind the scenes (or have it already latent in any new systems). They are already available for most languages and systems, architectures, and existing systems infrastructure tools, that could be used from office tools, or personal projects, to enterprise systems.

The question still remains: will these advances benefit us and create a better world, by respecting all the current and future legal and ethical initiatives of different organisms (and also common sense and human values), or are we creating the system of systems, that at some point in the future might become the best thing, or the perfect ultimate efficient thing replacing everything else, or cumbersome, or unusable, or dangerous, or worse ?

You can find more about these protocols at their official pages:

• MCP — https://modelcontextprotocol.io/ — “MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.”
• A2A — https://a2aproject.github.io/A2A/latest/ — “The Agent2Agent (A2A) Protocol is an open standard designed to enable seamless communication and collaboration between AI agents.”
• OCPI — https://evroaming.org/ocpi/ — The Open Charge Point Interface protocol (OCPI) supports connections between eMobility Service Providers (MSP) who have EV drivers as customers, and Charge Point Operators (CPO) who manage charge stations, used for Roaming.
• OCM — https://openchargemap.org/ — Protocol, map and ‘open data’ registry of electric vehicle charging locations.
• OCHP — https://github.com/e-clearing-net/OCHP — older protocol, it is featured and implemented here as example of using SOAP (wsdl schemas), to implement MCP Server.
• OSCP — older protocol featured here because the most complex newer alternatives and implementations require more work on developing production ready scale SDKs with MCP Servers.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI