A2A has moved beyond its 2025 launch announcement, but the evidence supports a more precise conclusion than “150 production customers.” In an April 9, 2026 announcement, the Linux Foundation said more than 150 organizations supported the Agent2Agent Protocol, up from more than 50 at launch. It also reported a stable A2A v1.0 release, integrations across Google, Microsoft and AWS ecosystems, and production deployments in sectors including supply chain, financial services, insurance and IT operations.
Those developments make A2A a credible open protocol project with substantial industry backing. They do not independently verify 150 mission-critical deployments, common traffic volumes, reliability results or seamless interoperability across every participating product.
What A2A is designed to solve
The Agent2Agent Protocol, or A2A, is an open standard for communication between independent AI agents. It is intended to let agents discover capabilities, exchange tasks and goals, report status, return results and delegate work across vendors, frameworks, clouds and organizational boundaries.
That is different from a conventional API integration. A traditional API usually exposes fixed application functions such as “create invoice” or “retrieve inventory.” A2A is designed for a higher-level interaction in which one agent can ask another agent to perform a task, negotiate capabilities and receive progress or an eventual result without exposing the remote agent’s internal memory, prompts or implementation.
The practical problem is fragmentation. Without a shared protocol, every agent pair may require a custom connector, authentication design, payload format and monitoring approach. That increases vendor lock-in and makes cross-cloud or cross-company delegation harder to secure and operate.
Google announced A2A on April 9, 2025, initially with support from more than 50 enterprise technology companies. Google later donated the project to the Linux Foundation on June 23, 2025, creating a more vendor-neutral governance model. Google’s anniversary account documents that timeline.
What the 150-organization milestone actually means
The Linux Foundation said on April 9, 2026, that more than 150 organizations supported A2A, compared with more than 50 at launch and more than 100 by June 2025. It named AWS, Cisco, Google, IBM, Microsoft, Salesforce, SAP and ServiceNow among the supporters. Read the Linux Foundation announcement.
“Supporting A2A” is an ecosystem metric, not a standardized production-customer metric. It may include organizations that contribute code, test the specification, build SDKs, provide tooling, participate in governance or publicly endorse the project.
The figure does not tell readers:
- How many organizations run A2A in production.
- How many implementations are certified or interoperable under the same test suite.
- How many agents, requests, workflows or users are involved.
- Whether deployments handle customer-facing or only internal workloads.
- Whether the implementations are generally available, pilots or demonstrations.
The milestone is meaningful evidence of industry interest and growing standardization pressure. It should not be presented as proof that 150 companies are operating large, mission-critical A2A systems.
How Google, Microsoft and AWS are adopting A2A
Google Cloud
Google created A2A and continues to position it as part of a broader agent ecosystem. Its partner announcement described agents communicating and coordinating across enterprise platforms such as Salesforce, SAP, ServiceNow, Workday, Atlassian, Box and Intuit. Google Cloud’s announcement describes ecosystem participation and contributions.
That does not mean every named partner offers a generally available, production-grade A2A implementation. Buyers should verify product availability, supported protocol version, geographic coverage, identity integration and operational support for each service.
Microsoft Foundry and Copilot Studio
Microsoft announced plans in May 2025 to support A2A in Azure AI Foundry and Copilot Studio. Its announcement described cross-cloud agent collaboration alongside controls involving Microsoft Entra, mutual TLS, content safety, audit logs, governance and service-level objectives. Microsoft’s announcement used forthcoming and preview-oriented language, so availability should not be assumed solely from that article.
Free tools Windows power users keep installed
One-click scans. No signup required.
The 2026 Linux Foundation announcement says Microsoft integrated A2A into Azure AI Foundry and Copilot Studio. Organizations should still confirm the exact product tier, region, version and general-availability status. Microsoft describes Foundry as a platform for building, deploying and scaling AI applications and agents with security, governance, monitoring and open-protocol support.
AWS Bedrock AgentCore
The Linux Foundation identifies Amazon Bedrock AgentCore Runtime as AWS’s route for A2A support. AWS presents AgentCore as production infrastructure for agents built with different frameworks and models, including authentication, access control, tracing, observability and deployment capabilities.
AgentCore infrastructure does not validate every possible A2A architecture. Customers remain responsible for agent logic, permissions, data governance, evaluation, identity design and safeguards around consequential actions.
What A2A v1.0 changes
A2A v1.0, announced in March 2026, is presented by the project as its first stable, production-ready specification. The release adds or formalizes several capabilities important to enterprise deployments:
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match- Multiple protocol bindings: JSON over HTTP, gRPC and JSON-RPC.
- Version negotiation: Clients and agents can determine compatible protocol behavior.
- Multi-tenancy: A single endpoint can securely host multiple agents.
- Signed Agent Cards: Cryptographic verification can help establish the authenticity of an agent’s identity and metadata.
- Updated security flows: The specification provides more enterprise-oriented security patterns.
- Infrastructure compatibility: Its web-aligned, stateless and layered design is intended to work with gateways and load balancers.
- Flexible task updates: Polling, streaming and webhooks can support different task lifecycles.
The release is not a guarantee that agents themselves are secure, accurate or reliable. Signed Agent Cards can help verify identity and metadata, but they do not decide what an agent is authorized to do, which data it may access or whether a human must approve a transaction.
Migration from v0.3
Organizations should not assume that v1.0 is a transparent upgrade. The v1.0 announcement says interaction behavior includes breaking changes, while Agent Card evolution is backward-compatible and can advertise support for both v0.3 and v1.0. The A2A v1.0 announcement should be treated as the migration reference.
Mixed-version testing should cover authentication, task states, error handling, streaming, retries, webhooks, timeouts and any intermediary gateways. A client that can discover a server does not necessarily behave correctly across every version boundary.
A2A versus MCP
A2A and the Model Context Protocol, or MCP, address different layers:
Recommended Free Tools
| Protocol | Primary role | Typical connection |
|---|---|---|
| MCP | Connects an agent or AI application to tools, data and context | Agent to database, API, file system or business tool |
| A2A | Coordinates communication and delegation between independent agents | Agent to agent across teams, vendors or organizations |
A procurement workflow might use both. A procurement agent could use MCP to query inventory and an enterprise resource planning system. It could then use A2A to delegate supplier discovery to an external sourcing agent. After receiving candidates and evidence, it could use MCP again to update internal systems.
The A2A project describes the protocols as complementary, not substitutes. A2A’s v1.0 documentation provides that distinction.
What “enterprise production use” should mean
The Linux Foundation says A2A is being used in production in supply chain, financial services, insurance and IT operations. That is an important claim, but the announcement does not provide a complete set of named deployments, traffic volumes, uptime figures, latency targets or independent verification.
For an enterprise evaluation, ask:
- Does A2A handle live customer or business traffic, or only internal testing?
- Are humans approving every consequential action?
- Do agents cross company or cloud boundaries?
- What happens when the remote agent is unavailable or returns stale data?
- Are tasks idempotent, retryable and auditable?
- Are latency, uptime, error rates and recovery times measured?
- Can the organization revoke delegated authority quickly?
- Is there a fallback to a deterministic workflow or human operator?
The most accurate formulation is therefore: the Linux Foundation reports enterprise production use. Publicly available evidence supports growing adoption and platform integration more strongly than it supports a claim of independently proven mass deployment.
Important limitations and failure modes
Protocol compatibility is not semantic compatibility
Two agents can exchange valid A2A messages while disagreeing about units, business terms, permissions, currencies, customer identifiers or the meaning of “completed.” A protocol standardizes communication; it does not create a shared business ontology.
Authentication is not authorization
Even when an agent’s identity is verified, the receiving system must decide what that agent may do. Use identity providers, gateways, policy engines, scoped credentials, tool permissions and application-level checks. High-impact actions may require human approval, transaction limits and separation of duties.
Remote agents create new trust boundaries
A remote agent can return malicious or outdated information, misrepresent its capabilities, exceed latency budgets, become unavailable or expose sensitive context through payloads and logs. Practical controls include allowlists, signed metadata, short-lived credentials, rate limits, timeouts, circuit breakers, output validation and careful data minimization.
A2A is not a complete workflow engine
A2A describes agent communication. Long-running business processes may still require durable execution, state persistence, idempotency keys, retry policies, dead-letter handling, compensation actions, human escalation, audit trails and replay.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Delegation can increase cost
A multi-agent design may reduce custom integration work while increasing model calls, network traffic, serialization, validation, observability storage, cross-region transfer and human-review costs. Teams should compare the cost of protocol-based delegation with a simpler API, queue or workflow engine.
When adopting A2A makes sense
A2A is most compelling when agents are built by different teams, vendors or frameworks; workflows cross cloud or organizational boundaries; and agents need task delegation rather than simple tool invocation. It can also be valuable when an organization wants a common discovery and communication layer instead of maintaining a separate connector for every agent pair.
It may be unnecessary when there is only one agent, all tools are internal and stable, or a conventional workflow engine already solves the problem. A vendor’s “A2A support” may also mean only a demo, SDK adapter or roadmap commitment. Verify the implementation before making the protocol a core dependency.
Governance and what comes next
The project’s governance story is still evolving. On August 17, 2026, Axios reported that A2A was moving from the Linux Foundation’s broader portfolio into the Agentic AI Foundation, alongside MCP and related agentic-AI projects. Axios reported the governance update.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
That move is significant for ecosystem positioning, but it is not evidence of additional technical adoption by itself. The next meaningful signals will be repeatable interoperability testing, transparent implementation matrices, named customer results, security guidance, reliable SDKs and clear version-management practices.
Bottom line for enterprise buyers
A2A has crossed an important credibility threshold. A stable v1.0 specification, support from major cloud ecosystems and more than 150 reported supporting organizations show that agent interoperability is being treated as a serious infrastructure problem rather than a one-company experiment.
But the strongest evidence is for ecosystem adoption and platform support—not a publicly audited count of 150 production customers. Evaluate A2A as a protocol layer, then separately assess the reliability, security, economics and governance of the agents and managed platforms built on top of it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors
