interoperability

advancedspreading

Interoperability means that AI agents, programs, and data sources from different manufacturers can work together under common rules without separate conversion.

In plain terms

Think of a power socket: you can plug in a hairdryer from any manufacturer because the plug and socket are standardized. Interoperability is the same in the software world: systems understand each other even if made by different companies.

This is important for agents because an agent alone is of little value. It is useful when it can access your calendar, your email, the company registry, and request help from another agent if needed.

Without common rules, every agent would need to be adapted individually to every program, which is expensive and slow. With common rules, the connector needs to be built only once and can be used by any agent.

This is more than simple connection. Both parties must not only reach each other but also understand the request and response in the same way.

How it works

Developers agree on a common protocol, meaning the format in which the agent sends requests and the format in which it receives answers.
Tools and data sources provide machine-readable descriptions of themselves: what they can do, what data they expect, and what they return.
When connecting, the agent queries this description and from it knows when to use which tool.
If two agents work together, they introduce themselves to each other in a similar way, then pass on a task and send back a result.
Permissions and authentication must be checked separately at every step, because a shared language alone does not grant authorization.

An everyday example

You ask your AI assistant to organize a business trip. The agent checks your calendar, searches for flight tickets through a travel site's agent, and then coordinates with the accommodation booking system. This only works smoothly if these systems communicate according to common rules.

Who is affected

entrepreneur, executiveIf the purchased AI tools are not interoperable, you tie your company to a single vendor, and switching later becomes costly.

developerWith a standard connector, you write an integration once, and it works with multiple agents. → Code

employeeThis determines whether your AI assistant can access the programs you use for daily work, or if you must copy data between them. → Work & HR

lawyer, citizenIn interoperable systems, your data moves between multiple actors, so it is important to know who has access to what. → Law

A common misconception

MythIf two AI systems support the same standard, they will work together safely and flawlessly.

In realityA common standard only fixes the form of communication. It does not guarantee that the other party is reliable, understands the task well, or is authorized to access your data.

Limits and risks

Multiple competing standards exist, so a product being "standard" does not mean it is compatible with everything.
The more systems an agent connects to, the more points there are where it can be attacked or deceived with malicious instructions.
When agents from multiple companies work together, it is difficult to determine who is responsible in case of an error.
Vendors often add their own extensions to the standard, so in practice you may still end up tied to them.

Similar, but not the same

integration – Integration is the specific connection of two concrete systems, whereas interoperability is a general capability based on common rules that works with any compatible system.
compatibility – Compatibility means that two systems can coexist or one replaces the other, while interoperability means they actually exchange data and work together.
portability – Portability is about being able to move your data or program to another provider, not about systems working together while running.

Related concepts

Broader

standardization, AI ecosystem, open standard

Narrower

Model Context Protocol, Agent2Agent protokoll

On the edge

multi-agent system, Coordinating agents, vendor lock-in, AI-agent, data portability, API, tool use

A brief history

The term originates from computing and telecommunications, where for decades it has denoted that systems from different manufacturers can work together. In the world of AI agents, it came to the fore in November 2024 when Anthropic published the Model Context Protocol, followed in April 2025 by Google presenting the Agent2Agent protocol aimed at communication between agents.

Where things stand

At the end of 2026, most major AI providers support the best-known open protocols. Stewardship of the Model Context Protocol and the Agent2Agent protocol has moved to a vendor-neutral framework under the Linux Foundation. However, seamless and secure cooperation between agents from different companies is still not a given.

Status: 2026 November

In the news

Appears in 4 Agent Age stories; the curve shows weekly occurrence in the news magazine.

AI shopping agents raise fraud and security risks ahead of the holiday seasonNov 2, 2026 · Marketing
Payment networks build infrastructure for agentic shopping: Mastercard Agent Connect, Know-Your-AgentSep 8, 2026 · Marketing
Google's A2A standard joins the Agentic AI Foundation as agent interoperability edges closerAug 17, 2026 · Agent Age
Prior authorization data goes public as health systems test real-time checks through EpicAug 17, 2026 · Health

Also known as

interoperability · interoperability · interop · interoperability · collaboration capability · agent interoperability · interoperability between agents · agent interoperability

Sources

Model Context Protocol – official documentation

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.