KDCube
← AI Industry News

AWS Sunsets Bedrock Agents Classic as AgentCore Becomes the Path

AWS moved its original Bedrock Agents into maintenance mode as Bedrock Agents Classic, freezing its model catalog and closing it to new customers as of July 30, 2026, with AgentCore now the sanctioned runtime. Meanwhile Insygna shipped a free pre-deployment agent security score and 6sense laun...

AI Industry Dossier — as of 2026-08-12

Highlights

  • AWS quietly moved its original Bedrock Agents — launched November 2023 — into maintenance mode, renaming it "Bedrock Agents Classic," freezing its model catalog, and closing it to new customers as of July 30, 2026; AgentCore is now the sanctioned path forward (AWS Docs)
  • Agent-security startup Insygna launched a free Agent Report Card that scores any agent across six security dimensions before it touches company systems, handing risk teams a pre-deployment gate instead of a post-incident autopsy (AIThority)
  • 6sense shipped an MCP server exposing its GTM intelligence — buying-stage predictions, 6QA status, keyword intent — callable from Claude, ChatGPT, Writer, and Agentforce with no custom integration (6sense)

Key Signals

  1. AWS retires its flagship agent product — the runtime, not the framework, wins July 30, 2026 (docs), surfaced this week

    Bedrock Agents Classic keeps running for existing workloads with no end-of-life date, but the model catalog is frozen at the maintenance-mode boundary and new models ship only through AgentCore (AWS Docs). Coming right after yesterday's AgentCore 14-day-session GA, the message is blunt: AWS is consolidating everyone onto the framework-agnostic runtime and letting the managed, opinionated agent builder wither. Teams still on Classic should treat migration planning as scheduled work, not optional.

  2. Agent governance moves left — to a pre-deployment score Aug 11, 2026

    Insygna's Agent Report Card lets any org connect an agent repo and get back per-agent scores, a findings list with version history, and an "Insygna Verified" badge, tied to a broader agentic-workforce identity layer spanning Teams, Slack, Copilot, Claude, and ChatGPT (AIThority; AI Agent News). After a fortnight of runtime-authority incidents and MCP-server vulnerability counts, "prove the agent is safe before it gets credentials" is becoming the buying question — and a free scan lowers the bar to asking it.

  3. MCP becomes the enterprise-data on-ramp, not just a tool protocol Aug 2026

    6sense's MCP server turns live buying signals into something any MCP-compatible agent can call, replacing manually exported target lists with up-to-date intent inside the agent workspace (6sense). It's a small launch with a big pattern: data vendors are now shipping MCP endpoints as the default distribution channel into agent runtimes, which is where connector-ecosystem gravity is heading.

Why It Matters / What To Watch

  1. Runtime lock-in is now a migration project, not a preference
    • If you built on Bedrock Agents Classic, inventory what depends on its frozen model catalog before a needed newer model forces the AgentCore port (AWS Docs).
    • Watch whether other vendors follow AWS in demoting managed agent builders in favor of lower-level runtimes — the abstraction that survives is the one that gets new models (AWS Docs).
  2. Pre-deployment agent scoring is the new procurement checkpoint
    • Evaluate whether a lightweight external score (six dimensions, shareable badge) fits your agent intake process before granting tool credentials (AIThority).
    • Track how agents ingest live enterprise data via MCP — vet each new MCP endpoint as an authority surface, not a convenience, given the connector's expanding blast radius (6sense; AI Agent News).

Quick Links