The A2A protocol (Agent2Agent), which lets AI agents from different vendors talk to each other, joined the Agentic AI Foundation (AAIF) on August 20, 2026. This is the same neutral Linux Foundation body that already hosts Anthropic's MCP protocol. For an SMB using or considering several AI agents (customer support, prospecting, accounting), this convergence addresses a real concern: staying locked into a single vendor.
In brief
- On August 20, 2026, the A2A protocol (created by Google) became a hosted project of the Agentic AI Foundation, the same neutral foundation that has managed Anthropic's MCP protocol since December 2025.
- A2A and MCP are complementary, not competing: MCP connects an AI agent to your tools (CRM, email, databases); A2A connects AI agents to each other, even across different vendors.
- More than 150 organizations already back A2A, including AWS, Cisco, Google, Microsoft, Salesforce, SAP and ServiceNow.
- The Agentic AI Foundation now counts more than 250 members, up from fewer than 40 at its creation in December 2025 (source: Agentic AI Foundation).
- In practice, this reduces the risk of single-vendor lock-in: AI agents from different brands will be able to collaborate without costly custom integration.
Why this announcement matters to your SMB
Many SMBs already use several AI tools without necessarily thinking of them this way: a customer support agent, a writing assistant, a prospecting tool, sometimes an AI module built into their ERP. Until now, making these agents work together when they come from competing vendors required custom technical integration, expensive to build and maintain.
A2A (Agent2Agent) is an open standard created by Google in April 2025, then handed to the Linux Foundation in June 2025. It lets an AI agent publish a "capability card": a description of what it can do, which other agents can read to delegate a task, without vendor-specific integration code. As Rao Surapaneni, VP at Google Cloud, put it: "A2A enables agents to easily discover, delegate, and collaborate across frameworks and siloed platforms, without custom integrations."
MCP and A2A do different things
MCP connects an AI agent to tools and data (vertically). A2A connects AI agents to each other, including across different companies (horizontally). An SMB may need both as it expands its AI use cases.
A shared governance to avoid vendor lock-in
The real change on August 20, 2026 is organizational, not technical. By joining the Agentic AI Foundation alongside MCP, A2A moves under neutral governance shared by direct competitors: AWS, Google and Microsoft sit together as founding members. Manik Surtani, CTO of AAIF, summarized the stakes: "A2A represents an important step toward an open, interoperable future for AI agents" (source: official Agentic AI Foundation blog).
April 2025
Google launches A2A
June 2025
A2A handed to the Linux Foundation
December 2025
Agentic AI Foundation is created
August 20, 2026
A2A joins AAIF
For an SMB, this shared governance is a trust signal: a standard controlled by a single vendor can change its rules overnight; a standard managed by a neutral foundation, with competitors at the table, offers more long-term stability.
MCP and A2A: two protocols, two purposes
| Criterion | MCP (Model Context Protocol) | A2A (Agent2Agent) |
|---|---|---|
| Created by | Anthropic (November 2024) | Google (April 2025) |
| Connects | An AI agent to tools and data | AI agents to each other |
| Direction | Vertical (agent to tool) | Horizontal (agent to agent) |
| Joined AAIF | December 2025 | August 2026 |
| Concrete example | Claude accesses your CRM | Your billing agent delegates a check to a supplier's agent |
| Adopted by | Claude, ChatGPT, Gemini, Copilot | Azure AI Foundry, AWS Bedrock, Google Cloud |
How an SMB can prepare for interoperable AI agents
Agent interoperability is still emerging: few SMBs have an immediate operational need today. But planning ahead avoids being forced into a technical choice later.
Map your current AI agents
Check MCP and A2A compatibility
Avoid closed proprietary integrations
Track progress without rushing
Don't overreact
A2A deployment currently remains concentrated on complex use cases at large enterprises (supply chain, finance, insurance, IT). An SMB generally doesn't need to act urgently, but benefits from factoring this criterion into future AI tool choices.
FAQ
What is the A2A protocol?
A2A (Agent2Agent) is an open standard created by Google in April 2025 that lets AI agents from different vendors discover each other, delegate tasks, and communicate without custom integration code.
What is the difference between MCP and A2A?
MCP connects an AI agent to tools and data (CRM, email, databases). A2A connects AI agents to each other, including across different companies and vendors. The two protocols are complementary.
Why did A2A join the same foundation as MCP?
To place both protocols under neutral, shared governance, with direct competitors (AWS, Google, Microsoft) as founding members. This prevents a central standard for agentic AI from depending on a single vendor's decisions.
Should an SMB adopt A2A-compatible tools right now?
Not urgently. Production deployment remains concentrated on complex large-enterprise use cases. Still, it's worth checking a tool's MCP/A2A compatibility when choosing a new AI vendor, as a longevity criterion.
Conclusion
This convergence between A2A and MCP under shared neutral governance marks a step toward a more open AI agent ecosystem, where an SMB would no longer need to buy everything from a single vendor for its AI tools to collaborate. Practical use is still emerging, but the signal sent by competitors like AWS, Google and Microsoft, aligned around the same standard, is worth following. To learn more about the protocols connecting your AI agents, see our guide to the MCP protocol update or browse all our AI resources for SMBs.
