News|
Weekly AI Agent Launch Roundup: ADK 2.0 Alpha, A2A 1.0, OpenAI Agents 0.13, and More
This week's AI agent launches for builders include Google ADK 2.0 alpha, A2A 1.0, OpenAI Agents 0.13, Mastra 1.14, CrewAI 1.11.1, and more.
Agent News Watch
Know what moved. Know what to build next.
Launches, protocol changes, and implementation guides filtered for engineering, product, and AI platform teams.
News
Latest dispatches
News|
Weekly AI Agent Launch Roundup: ADK 2.0 Alpha, A2A 1.0, OpenAI Agents 0.13, and More
This week's AI agent launches for builders include Google ADK 2.0 alpha, A2A 1.0, OpenAI Agents 0.13, Mastra 1.14, CrewAI 1.11.1, and more.
News|
A2A Protocol v1.0.0 Lands: What the Stable Agent-to-Agent Spec Changes for Builders
A2A v1.0.0 adds tasks/list, modern OAuth flows, multitenant gRPC support, and breaking spec cleanup. Here is what agent builders need to test.
News|
Google ADK 2.0 Alpha Adds a Workflow Runtime and Task API
Google ADK 2.0 alpha introduces graph-based workflow orchestration and structured task delegation. Here is what it changes for AI agent builders.
Guides
Implementation depth
Guide|
Multi-Agent Architecture: Patterns, Tradeoffs, and Reference Designs
Learn when multi-agent architecture outperforms single-agent systems, which coordination patterns fit best, and how to manage context, reliability, security, and cost.
Guide|
AI Agent Use Cases: Where Agents Actually Fit in Production
Learn the best AI agent use cases for product, ops, engineering, and support teams, plus how to choose the right autonomy level, architecture, and rollout path.
Guide|
AI Agent Examples: 12 Real Workflows and What They Teach Builders
Explore concrete AI agent examples across coding, research, support, operations, sales, and personal productivity, with tools, autonomy level, and build lessons.
Guide|
Agent-to-Agent Protocol: What A2A Is, How It Works, and How It Differs From MCP
Learn what Agent-to-Agent Protocol is, how A2A handles cross-agent communication, and when builders should care about A2A versus MCP.
Learning path
Start here
Topics
Browse by decision
Foundations
Start with definitions and real workflows
Separate real agent systems from chatbot theater, then study the jobs teams already automate well.
Start here before you score pilots, split roles, or shop stacks.
Pilot Selection
Choose the first workflow worth shipping
Turn examples into a real first sprint by scoring value, autonomy, risk, and approval needs.
Best when the question is what to test next, not what an agent is.
System Design
Map the control surfaces and specialist roles
Define the base architecture, decide when one agent is enough, and add orchestration only when the workflow needs it.
Use this lane once the pilot has to own state, retries, approvals, and handoffs.
Stack And Protocols
Choose the runtime, tools, and delegation layer
Compare frameworks, standardize capability access, and separate tool use from cross-agent delegation.
Use this lane when implementation details become the blocker.
Launch Controls
Ship with permissions, evaluation, and rollback paths
Attach rollout discipline to the build with tighter permissions, quality checks, and rollback paths.
Even strong stacks fail without security, measurement, and stop conditions.