Designing Collaborative Multi-Agent Techniques with the A2A Protocol – O’Reilly

It seems like each different AI announcement currently mentions “brokers.” And already, the AI neighborhood has 2025 pegged as “the yr of AI brokers,” generally with out far more element than “They’ll be wonderful!” Usually forgotten on this hype are the basics. All people is dreaming of armies of brokers, reserving resorts and flights, researching advanced matters, and writing PhD theses for us. And but we see little substance that addresses a crucial engineering problem of those bold techniques: How do these impartial brokers, constructed by completely different groups utilizing completely different tech, usually with fully opaque inside workings, truly collaborate?

However enterprises aren’t usually fooled by these hype cycles and guarantees. As an alternative, they have a tendency to chop by the noise and ask the onerous questions: If each firm spins up its personal intelligent agent for accounting, one other for logistics, a 3rd for customer support, and you’ve got your personal private assistant agent attempting to wrangle all of them—how do they coordinate? How does the accounting agent securely move information to the logistics agent with out a human manually copying knowledge between dashboards? How does your assistant delegate reserving a flight with no need to know the precise, proprietary, and sure undocumented inside workings of 1 explicit journey agent?

Proper now, the reply is usually “they don’t” or “with an entire lot of {custom}, brittle, painful integration code.” It’s turning into a digital Tower of Babel: Brokers get caught in their very own silos, unable to speak to one another. And with out that collaboration, they will’t ship on their promise of tackling advanced, real-world duties collectively.

The Agent2Agent (A2A) Protocol makes an attempt to handle these urgent questions. Its aim is to supply that lacking widespread language, a algorithm for a way completely different brokers and AI techniques can work together with no need to put open their inner secrets and techniques or get caught in custom-built, one-off integrations.

On this article, we’ll dive into the small print of A2A. We’ll have a look at:

• The core concepts behind it: What underlying rules is it constructed on?
• The way it truly works: What are the important thing mechanisms?
• The place it matches within the broader panorama, specifically, the way it compares to and doubtlessly enhances the Mannequin Context Protocol (MCP), which tackles the associated (however completely different) downside of brokers utilizing instruments.
• What we expect comes subsequent within the space of multi-agent system design.

A2A Protocol Overview

At its core, the A2A protocol is an effort to determine a means for AI brokers to speak and collaborate. Its goal is to supply a typical framework permitting brokers to:

• Uncover capabilities: Determine different out there brokers and perceive their features.
• Negotiate interplay: Decide the suitable modality for exchanging info for a particular job—easy textual content, structured kinds, even perhaps bidirectional multimedia streams.
• Collaborate securely: Execute duties cooperatively, passing directions and knowledge reliably and safely.

However simply itemizing objectives like “discovery” and “collaboration” on paper is simple. We’ve seen loads of bold tech requirements stumble as a result of they didn’t grapple with the messy realities early on (OSI community mannequin, anybody?). After we’re attempting to get numerous completely different techniques, constructed by completely different groups, to truly cooperate with out creating chaos, we want greater than a wishlist. We want some agency guiding rules baked in from the beginning. These mirror the hard-won classes about what it takes to make advanced techniques truly work: How can we deal with and make trade-offs in terms of safety, robustness, and sensible utilization?

With that in thoughts, A2A was constructed with these tenets:

• Easy: As an alternative of reinventing the wheel, A2A leverages well-established and broadly understood present requirements. This lowers the barrier to adoption and integration, permitting builders to construct upon acquainted applied sciences.
• Enterprise prepared: A2A consists of strong mechanisms for authentication (verifying agent identities), safety (defending knowledge in transit and at relaxation), privateness (guaranteeing delicate info is dealt with appropriately), tracing (logging interactions for auditability), and monitoring (observing the well being and efficiency of agent communications).
• Async first: A2A is designed with asynchronous communication as a major consideration, permitting duties to proceed over prolonged intervals and seamlessly combine human-in-the-loop workflows.
• Modality agnostic: A2A helps interactions throughout numerous modalities, together with textual content, bidirectional audio/video streams, interactive kinds, and even embedded iframes for richer person experiences. This flexibility permits brokers to speak and current info in essentially the most acceptable format for the duty and person.
• Opaque execution: It is a cornerstone of A2A. Every agent taking part in a collaboration stays invisible to the others. They don’t have to reveal their inner reasoning processes, their data illustration, reminiscence, or the precise instruments they could be utilizing. Collaboration happens by well-defined interfaces and message exchanges, preserving the autonomy and mental property of every agent. Be aware that, whereas brokers function this fashion by default (with out revealing their particular implementation, instruments, or mind-set), a person distant agent can select to selectively reveal elements of its state or reasoning course of by way of messages, particularly for UX functions, equivalent to offering person notifications to the caller agent. So long as the choice to disclose info is the duty of the distant agent, the interplay maintains its opaque nature.

Taken collectively, these tenets paint an image of a protocol attempting to be sensible, safe, versatile, and respectful of the impartial nature of brokers. However rules on paper are one factor; how does A2A truly implement these concepts? To see that, we have to shift from the design philosophy to the nuts and bolts—the precise mechanisms and parts that make agent-to-agent communication work.

Key Mechanisms and Parts of A2A

Translating these rules into observe requires particular mechanisms. Central to enabling brokers to know one another inside the A2A framework is the Agent Card. This element features as a standardized digital enterprise card for an AI agent, sometimes offered as a metadata file. Its major objective is to publicly declare what an agent is, what it will possibly do, the place it may be reached, and easy methods to work together with it.

Right here’s a simplified instance of what an Agent Card would possibly seem like, conveying the important info:

{
  "title": "StockInfoAgent",
  "description": "Gives present inventory worth info.",
  "url": "http://stock-info.instance.com/a2a",
  "supplier": { "group": "ABCorp" },
  "model": "1.0.0",
  "expertise": [
    {
      "id": "get_stock_price_skill",
      "name": "Get Stock Price",
      "description": "Retrieves current stock price for a company"
    }
  ]
}

(shortened for brevity)

The Agent Card serves as the important thing connector between the completely different actors within the A2A protocol. A consumer—which might be one other agent or maybe the applying the person is interacting with—finds the Agent Card for the service it wants. It makes use of the small print from the cardboard, just like the URL, to contact the distant agent (server), which then performs the requested job with out exposing its inner strategies and sends again the outcomes in line with the A2A guidelines.

As soon as brokers are in a position to learn one another’s capabilities, A2A constructions their collaboration round finishing particular duties. A job represents the basic unit of labor requested by a consumer from a distant agent. Importantly, every job is stateful, permitting it to trace progress over time, which is important for dealing with operations which may not be instantaneous—aligning with A2A’s “async first” precept.

Communication associated to a job primarily makes use of messages. These carry the continued dialogue, together with preliminary directions from the consumer, standing updates, requests for clarification, and even intermediate “ideas” from the agent. When the duty is full, the ultimate tangible outputs are delivered as artifacts, that are immutable outcomes like information or structured knowledge. Each messages and artifacts are composed of a number of elements, the granular items of content material, every with an outlined sort (like textual content or a picture).

This whole change depends on commonplace net applied sciences like HTTP and customary knowledge codecs, guaranteeing a broad basis for implementation and compatibility. By defining these core objects—job, message, artifact, and half—A2A offers a structured means for brokers to handle requests, change info, and ship outcomes, whether or not the work takes seconds or hours.

Safety is, after all, a crucial concern for any protocol aiming for enterprise adoption, and A2A addresses this instantly. Somewhat than inventing fully new safety mechanisms, it leans closely on established practices. A2A aligns with requirements just like the OpenAPI specification for outlining authentication strategies and usually encourages treating brokers like different safe enterprise functions. This enables the protocol to combine into present company safety frameworks, equivalent to established identification and entry administration (IAM) techniques for authenticating brokers, making use of present community safety guidelines and firewall insurance policies to A2A endpoints, or doubtlessly feeding A2A interplay logs into centralized safety info and occasion administration (SIEM) platforms for monitoring and auditing.

A core precept is conserving delicate credentials, equivalent to API keys or entry tokens, separate from the principle A2A message content material. Purchasers are anticipated to acquire these credentials by an impartial course of. As soon as obtained, they’re transmitted securely utilizing commonplace HTTP headers, a typical observe in net APIs. Distant brokers, in flip, clearly state their authentication necessities—usually inside their Agent Playing cards—and use commonplace HTTP response codes to handle entry makes an attempt, signaling success or failure in a predictable means. This reliance on acquainted net safety patterns lowers the barrier to implementing safe agent interactions.

A2A additionally facilitates the creation of a distributed “interplay reminiscence” throughout a multi-agent system by offering a standardized protocol for brokers to change and reference task-specific info, together with distinctive identifiers (taskId, sessionId), standing updates, message histories, and artifacts. Whereas A2A itself doesn’t retailer this reminiscence, it allows every taking part A2A consumer and server agent to keep up its portion of the general job context. Collectively, these particular person agent reminiscences, linked and synchronized by A2A’s structured communication, kind the excellent interplay reminiscence of the complete multi-agent system, permitting for coherent and stateful collaboration on advanced duties.

So, in a nutshell, A2A is an try and convey guidelines and standardization to the quickly evolving world of brokers by defining how impartial techniques can uncover one another, collaborate on duties (even long-running ones), and deal with safety utilizing well-trodden net paths, all whereas conserving their inside workings non-public. It’s targeted squarely on agent-to-agent communication, attempting to unravel the issue of remoted digital employees unable to coordinate.

However getting brokers to speak to one another is just one piece of the interoperability puzzle going through AI builders immediately. There’s one other commonplace gaining important traction that tackles a associated but distinct problem: How do these subtle AI functions work together with the surface world—the databases, APIs, information, and specialised features sometimes called “instruments”? This brings us to Anthropic’s Mannequin Context Protocol, or MCP.

MCP: Mannequin Context Protocol Overview

It wasn’t so way back, actually, that giant language fashions (LLMs), whereas spectacular textual content mills, had been usually mocked for his or her generally hilarious blind spots. Requested to do easy arithmetic, depend the letters in a phrase precisely, or inform you the present climate, and the outcomes might be confidently delivered but fully fallacious. This wasn’t only a quirk; it highlighted a basic limitation: The fashions operated purely on the patterns discovered from their static coaching knowledge, disconnected from reside info sources or the flexibility to execute dependable procedures. However nowadays are principally over (or so it appears)—state-of-the-art AI fashions are vastly more practical than their predecessors from only a yr or two in the past.

A key purpose for the effectiveness of AI techniques (brokers or not) is their potential to attach past their coaching knowledge: interacting with databases and APIs, accessing native information, and using specialised exterior instruments. Equally to interagent communication, nonetheless, there are some onerous challenges that should be tackled first.

Integrating these AI techniques with exterior “instruments” includes collaboration between AI builders, agent architects, device suppliers, and others. A major hurdle is that device integration strategies are sometimes tied to particular LLM suppliers (like OpenAI, Anthropic, or Google), and these suppliers deal with device utilization in a different way. Defining a device for one system requires a particular format; utilizing that very same device with one other system usually calls for a distinct construction.

Take into account the next examples.

OpenAI’s API expects a operate definition structured this fashion:

{
  "sort": "operate",
  "operate": {
    "title": "get_weather",
    "description": "Retrieves climate knowledge ...",
    "parameters": {...}
  }
}

Whereas Anthropic’s API makes use of a distinct structure:

{
  "title": "get_weather",
  "description": "Retrieves climate knowledge ...",
  "input_schema": {...}
}

This incompatibility means device suppliers should develop and preserve separate integrations for every AI mannequin supplier they need to help. If an agent constructed with Anthropic fashions wants sure instruments, these instruments should observe Anthropic’s format. If one other developer desires to make use of the identical instruments with a distinct mannequin supplier, they basically duplicate the mixing effort, adapting definitions and logic for the brand new supplier.

Format variations aren’t the one problem; language limitations additionally create integration difficulties. For instance, getting a Python-based agent to instantly use a device constructed round a Java library requires appreciable improvement effort.

This integration problem is exactly what the Mannequin Context Protocol was designed to unravel. It gives a typical means for various AI functions and exterior instruments to work together.

Much like A2A, MCP operates utilizing two key elements, beginning with the MCP server. This element is liable for exposing the device’s performance. It comprises the underlying logic—possibly Python code hitting a climate API or routines for knowledge entry—developed in an appropriate language. Servers generally bundle associated capabilities, like file operations or database entry instruments. The second element is the MCP consumer. This piece sits contained in the AI utility (the chatbot, agent, or coding assistant). It finds and connects to MCP servers which might be out there. When the AI app or mannequin wants one thing from the surface world, the consumer talks to the fitting server utilizing the MCP commonplace.

The bottom line is that communication between consumer and server adheres to the MCP commonplace. This adherence ensures that any MCP-compatible consumer can work together with any MCP server, irrespective of the consumer’s underlying AI mannequin or the language used to construct the server.

Adopting this commonplace gives a number of benefits:

• Construct as soon as, use wherever: Create a functionality as an MCP server as soon as; any MCP-supporting utility can use it.
• Language flexibility: Develop servers within the language finest suited to the duty.
• Leverage ecosystem: Use present open supply MCP servers as a substitute of constructing each integration from scratch.
• Improve AI capabilities: Simply give brokers, chatbots, and assistants entry to numerous real-world instruments.

Adoption of MCP is accelerating, demonstrated by suppliers equivalent to GitHub and Slack, which now provide servers implementing the protocol.

MCP and A2A

However how do the Mannequin Context Protocol and the Agent2Agent (A2A) Protocol relate? Do they remedy the identical downside or serve completely different features? The strains can blur, particularly since many agent frameworks enable treating one agent as a device for one more (agent as a device).

Each protocols enhance interoperability inside AI techniques, however they function at completely different ranges. By analyzing their variations in implementation and objectives we will clearly establish key differentiators.

MCP focuses on standardizing the hyperlink between an AI utility (or agent) and particular, well-defined exterior instruments or capabilities. MCP makes use of exact, structured schemas (like JSON Schema) to outline instruments, establishing a transparent API-like contract for predictable and environment friendly execution. For instance, an agent needing the climate would use MCP to name a get_weather device on an MCP climate server, specifying the placement “London.” The required enter and output are strictly outlined by the server’s MCP schema. This strategy removes ambiguity and solves the issue of incompatible device definitions throughout LLM suppliers for that particular operate name. MCP normally includes synchronous calls, supporting dependable and repeatable execution of features (except, after all, the climate in London has modified within the meantime, which is fully believable).

A2A, then again, standardizes how autonomous brokers talk and collaborate. It excels at managing advanced, multistep duties involving coordination, dialogue, and delegation. Somewhat than relying on inflexible operate schemas, A2A interactions make the most of pure language, making the protocol higher suited to ambiguous objectives or duties requiring interpretation. A great instance could be “Summarize market traits for sustainable packaging.” Asynchronous communication is a key tenet of A2A, which additionally consists of mechanisms to supervise the lifecycle of probably prolonged duties. This includes monitoring standing (like working, accomplished, and enter required) and managing the required dialogue between brokers. Take into account a trip planner agent utilizing A2A to delegate book_flights and reserve_hotel duties to specialised journey brokers whereas monitoring their standing. In essence, A2A’s focus is the orchestration of workflows and collaboration between brokers.

This distinction highlights why MCP and A2A operate as complementary applied sciences, not opponents. To borrow an analogy: MCP is like standardizing the wrench a mechanic makes use of—defining exactly how the device engages with the bolt. A2A is like establishing a protocol for a way that mechanic communicates with a specialist mechanic throughout the workshop (“Listening to a rattle from the entrance left, are you able to diagnose?”), initiating a dialogue and collaborative course of.

In subtle AI techniques, we will simply think about them working collectively: A2A would possibly orchestrate the general workflow, managing delegation and communication between completely different brokers, whereas these particular person brokers would possibly use MCP beneath the hood to work together with particular databases, APIs, or different discrete instruments wanted to finish their a part of the bigger job.

Placing It All Collectively

We’ve mentioned A2A for agent collaboration and MCP for device interplay as separate ideas. However their actual potential would possibly lie in how they work collectively. Let’s stroll by a easy, sensible state of affairs to see how these two protocols may operate in live performance inside a multi-agent system.

Think about a person asks their major interface agent—let’s name it the Host Agent—a simple query: “What’s Google’s inventory worth proper now?”

The Host Agent, designed for person interplay and orchestrating duties, doesn’t essentially know easy methods to fetch inventory costs itself. Nonetheless, it is aware of (maybe by consulting an agent registry by way of an Agent Card) a few specialised Inventory Data Agent that handles monetary knowledge. Utilizing A2A, the Host Agent delegates the duty: It sends an A2A message to the Inventory Data Agent, basically saying, “Request: Present inventory worth for GOOGL.”

The Inventory Data Agent receives this A2A job. Now, this agent is aware of the precise process to get the information. It doesn’t want to debate it additional with the Host Agent; its job is to retrieve the value. To do that, it turns to its personal toolset, particularly an MCP inventory worth server. Utilizing MCP, the Inventory Data Agent makes a exact, structured name to the server—successfully get_stock_price(image: "GOOGL"). This isn’t a collaborative dialogue just like the A2A change; it’s a direct operate name utilizing the standardized MCP format.

The MCP server does its job: appears up the value and returns a structured response, possibly {"worth": "174.92 USD"}, again to the Inventory Data Agent by way of MCP.

With the information in hand, the Inventory Data Agent completes its A2A job. It sends a ultimate A2A message again to the Host Agent, reporting the outcome: "Outcome: Google inventory is 174.92 USD."

Lastly, the Host Agent takes this info obtained by way of A2A and presents it to the person.

Even on this easy instance, the complementary roles develop into clear. A2A handles the higher-level coordination and delegation between autonomous brokers (Host delegates to Inventory Data). MCP handles the standardized, lower-level interplay between an agent and a particular device (Inventory Data makes use of the value server). This creates a separation of considerations: The Host agent doesn’t have to find out about MCP or inventory APIs, and the Inventory Data agent doesn’t have to deal with advanced person interplay—it simply fulfills A2A duties, utilizing MCP instruments the place obligatory. Each brokers stay largely opaque to one another, interacting solely by the outlined protocols. This modularity, enabled through the use of each A2A for collaboration and MCP for device use, is vital to constructing extra advanced, succesful, and maintainable AI techniques.

Conclusion and Future Work

We’ve outlined the challenges of constructing AI brokers collaborate, explored Google’s A2A protocol as a possible commonplace for interagent communication, and in contrast and contrasted it with Anthropic’s Mannequin Context Protocol. Standardizing device use and agent interoperability are vital steps ahead in enabling efficient and environment friendly multi-agent system (MAS) design.

However the story is way from over, and agent discoverability is without doubt one of the quick subsequent challenges that should be tackled. When speaking to enterprises it turns into obviously apparent that that is usually very excessive on their precedence listing. As a result of, whereas A2A defines how brokers talk as soon as related, the query of how they discover one another within the first place stays a major space for improvement. Easy approaches will be carried out—like publishing an Agent Card at a typical net tackle and capturing that tackle in a listing—however that feels inadequate for constructing a really dynamic and scalable ecosystem. That is the place we see the idea of curated agent registries come into focus, and it’s maybe some of the thrilling areas of future work for MAS.

We think about an inner “agent retailer” (akin to an app retailer) or skilled itemizing for a company’s AI brokers. Builders may register their brokers, full with versioned expertise and capabilities detailed of their Agent Playing cards. Purchasers needing a particular operate may then question this registry, looking not simply by title however by required expertise, belief ranges, or different important attributes. Such a registry wouldn’t simply simplify discovery; it could foster specialization, allow higher governance, and make the entire system extra clear and manageable. It strikes us from merely discovering an agent to discovering the proper agent for the job based mostly on its declared expertise.

Nonetheless, even subtle registries can solely assist us discover brokers based mostly on these declared capabilities. One other fascinating, and maybe extra basic, problem for the long run: coping with emergent capabilities. One of many outstanding elements of recent brokers is their potential to mix numerous instruments in novel methods to deal with unexpected issues. An agent outfitted with numerous mapping, visitors, and occasion knowledge instruments, as an example, might need “route planning” listed on its Agent Card. However by creatively combining these instruments, it may additionally be able to producing advanced catastrophe evacuation routes or extremely customized multistop itineraries—essential capabilities probably unlisted just because they weren’t explicitly predefined. How can we reconcile the necessity for predictable, discoverable expertise with the highly effective, adaptive problem-solving that makes brokers so promising? Discovering methods for brokers to sign or for shoppers to find these unlisted prospects with out sacrificing construction is a major open query for the A2A neighborhood and the broader area (as highlighted in discussions like this one).

Addressing this problem provides one other layer of complexity when envisioning future MAS architectures. Wanting down the highway, particularly inside giant organizations, we’d see the registry thought evolve into one thing akin to the “knowledge mesh” idea—a number of, doubtlessly federated registries serving particular domains. This might result in an “agent mesh”: a resilient, adaptable panorama the place brokers collaborate successfully beneath a unified centralized governance layer and distributed administration capabilities (e.g., introducing notions of an information/agent steward who manages the standard, accuracy, and compliance of a enterprise unit knowledge/brokers). However guaranteeing this mesh can leverage each declared and emergent capabilities shall be key. Exploring that totally, nonetheless, is probably going a subject for one more day.

Finally, protocols like A2A and MCP are important constructing blocks, however they’re not the complete map. To construct multi-agent techniques which might be genuinely collaborative and strong, we want extra than simply commonplace communication guidelines. It means stepping again and pondering onerous in regards to the total structure, wrestling with sensible complications like safety and discovery (each the express form and the implicit, emergent type), and acknowledging that these requirements themselves must adapt as we study. The journey from immediately’s often-siloed brokers to really cooperative ecosystems is ongoing, however initiatives like A2A provide useful markers alongside the best way. It’s undoubtedly a troublesome engineering highway forward. But, the prospect of AI techniques that may actually work collectively and deal with advanced issues in versatile methods? That’s a vacation spot well worth the effort.