Vertex AI Agent Builder GA; MCP Vendor Wave Hits With 21 Days to Go

<p><strong>Google promoted Vertex AI Agent Builder to GA</strong> with A2A protocol support and a built-in eval harness, completing a three-way managed agent runtime market alongside Bedrock AgentCore and Microsoft Agent Framework v1. Meanwhile, Stripe, GitHub, and Atlassian shipped stateless-ready MCP adapters ahead of the July 28 deadline, and <strong>L...

Platform Intel · 2026-07-07

Highlights

  • Google promoted Vertex AI Agent Builder to general availability with native A2A protocol support and a built-in eval harness, putting a third enterprise-grade managed agent runtime on the board alongside Amazon Bedrock AgentCore and Microsoft Agent Framework v1 (Google Cloud)
  • With 21 days until the MCP July 28 stateless-transport deadline, Stripe, GitHub, and Atlassian each shipped stateless-ready connector adapters this week, each changelog explicitly documenting the six breaking changes from the RC (MCP Spec)
  • Langfuse v3.1 introduced GPAI-aligned trace-export schemas, letting teams generate EU-compliant audit records directly from existing LangGraph or Hermes instrumentation — 26 days before August 2 enforcement (Langfuse)
  • CrewAI v0.90 shipped persistent crew memory backed by configurable vector stores, closing its main gap with LangGraph’s PostgreSQL checkpoint adapters that landed last week (CrewAI)

Key Signals

  1. July 7, 2026
    Vertex AI Agent Builder reaches GA with A2A and eval

    Google’s promotion to GA adds first-class Agent-to-Agent (A2A) protocol support and a native evaluation harness runnable directly in the Vertex console without exporting to a separate eval platform. The built-in eval and managed runtime position it as a direct enterprise alternative to Bedrock AgentCore for teams already in GCP. (Google Cloud)

  2. Week of July 7
    MCP vendor ecosystem catches up with 21 days to spare

    Stripe, GitHub, and Atlassian each published production-ready stateless-transport adapters for their MCP connectors, with migration guides calling out the six breaking changes in the July 28 RC by name. Operators should audit transport negotiation and session-resumption logic this week — the final spec drops July 28 with no grace period. (MCP Spec)

  3. July 6–7, 2026
    Langfuse 3.1 ships GPAI-ready trace schemas

    Langfuse v3.1 adds structured trace-export templates aligned to EU GPAI documentation requirements. Teams can produce compliant audit records directly from existing agent instrumentation without post-processing. Langfuse 3.1 also accounts for the new session boundary semantics introduced by the MCP July 28 RC’s stateless transport changes. (Langfuse)


Why It Matters / What To Watch

  1. The managed multi-agent platform race is now three-way — pick your stack this month
    • Google Vertex GA, Amazon Bedrock AgentCore, and Microsoft Agent Framework v1 are all production-ready simultaneously for the first time. Pricing, regional coverage, and native eval depth differ significantly; teams without a committed platform should run comparative evals now before lock-in deepens. (Google Cloud)
    • Vertex’s A2A support out of the box means hybrid deployments mixing Google-managed agents with self-hosted LangGraph or CrewAI orchestration are now straightforward — worth testing if you’re on GCP but running your own orchestration layer. (Google Cloud)
  2. July 28 and August 2 deadlines are converging — prioritize the MCP trace model
    • MCP stateless transport changes how sessions terminate and resume, directly affecting trace completeness. Teams updating connectors for the July 28 deadline should validate that their observability layer captures the new session boundary events — Langfuse 3.1’s GPAI schemas are the fastest path to closing both gaps at once. (Langfuse)
    • The Stripe, GitHub, and Atlassian migration guides are the clearest practical reference for the six breaking changes. Teams running custom or internal MCP servers should diff against those guides before July 21 to leave time for staging validation. (MCP Spec)

Quick Links