Cutting through the noise in AI

Welcome to the Women Who AI Newsletter, your weekly update on what actually matters in AI when you’re focused on building and scaling startups.

Was this forwarded to you? We send out this newsletter every Monday morning. Click here to subscribe and join our community of founders shaping the future of AI.

This Week’s Deep Dive: A2A Protocol - A New Communication Standard for AI Agents

A2A is an open protocol developed by Google that provides a standard way for agents to collaborate, regardless of the underlying framework or vendor. Think of it as a universal translator for AI agents, allowing them to discover each other's capabilities, securely exchange information, and coordinate actions across enterprise platforms or applications.

At its core, A2A facilitates communication between a "client" agent and a "remote" agent through four key capabilities:

1. Capability discovery: Agents advertise their capabilities using an "Agent Card" in JSON format, allowing them to identify partners with complementary skills.

2. Task management: Communication between agents is oriented towards task completion, with support for both immediate and long-running tasks.

3. Collaboration: Agents can send each other messages to communicate context, replies, artifacts, or user instructions.

4. User experience negotiation: Agents can negotiate the correct format needed for responding to users, whether that's text, images, forms, or video.

Example Walk-Through

Imagine you're building a productivity suite with three specialized agents:

1. A scheduling agent built with LangChain

2. A document summarization agent built with Google's ADK

3. A data analysis agent built with CrewAI

Without a common protocol, integrating these would require custom connectors between each pair of agents—a brittle and maintenance-heavy approach.

With A2A, here's how they would interact:

1. When a user asks, "Summarize last quarter's sales report and schedule a team meeting to discuss it," the main agent recognizes this requires multiple specialized agents.

2. It queries the A2A "Agent Cards" of available agents to discover their capabilities.

3. It sends the document to the summarization agent using A2A's task management framework, which handles the document part using FilePart in the protocol.

4. The summarization agent processes the document and returns a summary as a TextPart.

5. The main agent then passes relevant data to the analysis agent, which identifies key trends.

6. Finally, it instructs the scheduling agent to create a meeting with the summary and analysis attached.

All this happens seamlessly using standardized messages, with each agent potentially running on different frameworks or even different vendors' platforms.

How is this different than MCP?

According to Google's official stance, A2A and MCP (Model Context Protocol) are complementary rather than competitive. The recommendation is to use "MCP for tools and A2A for agents."

MCP focuses on standardizing how applications provide context to LLMs, helping integrate legacy data systems and APIs with LLM-based applications. A2A, meanwhile, focuses on inter-agent communication with features MCP lacks: secure collaboration, task and state management, user experience negotiation, and capability discovery.

While they solve different immediate problems, there is some overlapping territory. We're moving beyond single, monolithic AI models towards complex ecosystems of specialized AI agents, and the protocol that enables these agents to interact effectively will become increasingly important.

Use Cases for Founders: Automation

• Multi-vendor business workflows: A2A enables seamless candidate sourcing across multiple specialized agents handling different aspects like finding candidates, scheduling interviews, and conducting background checks—even if these agents are built by different vendors.

• Supply chain optimization: Create agent systems that monitor inventory, coordinate with vendor agents, optimize logistics, and automatically adjust orders based on real-time data—with each specialized function potentially handled by different agent frameworks best suited to that particular task.

Use Cases for Founders: Building with A2A

• Agent marketplaces: A2A defines a standardized "Agent Card" system (typically hosted at /.well-known/agent.json) that lets agents advertise their capabilities. This creates opportunities for agent discovery platforms and marketplaces where specialized agents can be found and integrated into workflows.

• Composable Feature Pods: Build your product as a collection of micro-agents (e.g., “Billing Agent,” “Support Agent”) that can be swapped or upgraded independently.

Links to Dive Deeper

• Google’s official A2A announcement

• Tutorial: Getting Started with A2A

Hackathons

Ready to build that product you've been dreaming about? Check out these upcoming hackathons!

If you'd like to find a Women Who AI team for any event, reply to this email, and we'll connect everyone interested.

MCP and A2A Hackathon - AWS Edition | SF | Friday, May 2nd

Sundai HumanX Education Hack | Boston | Sunday, May 4th

Perplexity Hackathon | Virtual | Today-May 28

Share the Knowledge

If you found value in today's newsletter, please consider forwarding it to other women in your network who are building, or thinking about building, in the AI space.

Here's to building the future of AI, together.

Lea & Daniela