[1/x] Model Context Protocol (MCP) and Agent-to-Agent (A2A) (AI column)
In preparation for report I decided to study how we came to two of the most popular protocols for LLM. And I'm talking about Anthropic's MCP and Google's A2A protocol. Now it seems that we are moving towards a future where AI assistants of all kinds freely connect to the data they need. (via MCP) And they come together to solve our problems. (via A2A). But it wasn't always that way, so let's take a step back and see how these protocols came about.
Model Context Protocol (MCP) It is an open standard developed by Anthropic and represented 25 November 2024 years. Anthropic announced MCP as a solution to connect AI assistants to external systems and data, in order to overcome the fragmentation of integrations, since before that it was necessary to write custom integrations with each data source. The launch of MCP caught the attention of the industry: Anthropic opened the source code of the specification and SDK protocoland also provided ready-made server connectors for popular systems (Google Drive, Slack, GitHub, databases, etc.)
- In March. 2025 OpenAI has officially announced the introduction of MCP in all its products. Google DeepMind also announced support for MCP in April. 2025 Demis Hassabis has confirmed that upcoming Gemini models will be compatible with MCP to connect to external data. Eventually, by the middle 2025 MCP has become a de facto industry standard for data integration for LLM assistants, supported by OpenAI products. (including ChatGPT)Anthropic Claude, Microsoft platforms (Semantic Kernel, Azure OpenAI) and others. The media dubbed MCP a kind of “USB-C for AI” – a single connector that unites competitors for compatibility.
Agent-to-Agent (A2A) It is an open messaging protocol between AI agents that has been presented Google 9 April 2025 For more than a year on the Google Cloud platform 50 partners. The A2A initiative arose from the need to enable multiple autonomous agents to interact in enterprise scenarios. In its official announcement, Google noted that enterprises are increasingly implementing agent solutions to automate workflows and such agents must be able to collaborate with each other, even if they are created by different vendors or on different platforms. The A2A protocol was conceived as a universal language for such agents, allowing them to discover each other, share information and coordinate activities within a distributed ecosystem.
The launch of the A2A immediately received broad support in the industry. In addition to Google, Atlassian, Box, Cohere, Intuit, LangChain, MongoDB, PayPal, Salesforce, SAP, Accenture, Deloitte, IBM and others participated in the work on the standard.
- June. 2025 Google pass on A2A for the Linux Foundation for Neutral Community Development (It's kind of like Google. pass on Kubernetes is part of the Cloud Native Computing Foundation.) At the same time, it was announced that such giants as AWS, Cisco, Microsoft, Salesforce, SAP, etc. were participating in the A2A project. In the middle 2025 A2A was established as an open cross-platform standard supported by a host of major players, just as MCP became the standard for data connectivity.
Next Next post posture It will be about the MCP protocol.
#AI #Engineering #Architecture #DistributedSystems #ML #Software