AWS Just Killed Bedrock Agents. Here Is What AgentCore Means for Your AI Workloads
If You Built on Bedrock Agents, Your Clock Is Ticking
Amazon just made it official: Bedrock Agents, the service launched in November 2023, is now Bedrock Agents Classic. No new customers after July 30, 2026. Existing customers get maintenance mode. The replacement is Amazon Bedrock AgentCore.
If you built production AI agents on Bedrock Agents, you have a migration to plan. If you were about to start building, your entry point just changed. Here is what you need to know.
What Is Bedrock AgentCore?
AgentCore is not a renamed Bedrock Agents. It is a complete re-architecture built around five pillars:
| Component | What It Does |
|---|---|
| AgentCore Runtime | Deploy and operate agents at scale with memory, identity, and tool integration |
| AgentCore Gateway | Centralized tool server with security guard, translation, and composition |
| AgentCore Memory | Persistent context across sessions without rebuilding state |
| AgentCore Evaluations | Measure accuracy, performance, and reliability in dev and production |
| AgentCore Multi-Agent | Orchestrate multiple agents working together on complex tasks |
The old Bedrock Agents was a single-agent framework with basic tool calling. AgentCore lets you build systems of agents that share memory, delegate tasks, and route through a unified gateway.
What Changes for Existing Bedrock Agents Users
Immediate:
- No new customer sign-ups after July 30, 2026
- Existing agents continue to run in maintenance mode
- No new features for Bedrock Agents Classic
Migration path:
- AWS provides the AgentCore Import-Agent toolkit on GitHub
- Converts existing Bedrock Agents to LangChain, LangGraph, or Strands frameworks
- AgentCore supports the same Bedrock foundation models plus custom SageMaker models
What breaks:
- Custom tool configurations need rewriting for Gateway
- Memory state does not transfer automatically
- IAM roles and permissions need reconfiguration for the multi-agent model
Architecture Comparison
| Bedrock Agents Classic | Bedrock AgentCore | |
|---|---|---|
| Agent model | Single agent per application | Multi-agent orchestration |
| Memory | Session-scoped | Persistent cross-session |
| Tool integration | Direct API calls | Gateway with translation layer |
| Evaluation | Manual testing | Built-in evaluation suite |
| Model support | Bedrock models only | Bedrock + SageMaker custom models |
| Scaling | Per-agent | Infrastructure-managed auto-scale |
What This Means for Your AWS Bill
AgentCore pricing is per-agent invocation plus Gateway request charges. The Gateway adds a translation layer that normalizes tool calls across agents, which means fewer custom integrations to maintain but more AWS service charges on your bill.
If you run managed AWS through a partner like ServerGurus, your cost optimization should include Gateway usage analysis. Translation requests are metered. Composition calls that chain multiple tools count as separate invocations. A poorly architected multi-agent setup can 3x your AgentCore bill compared to a single-agent Bedrock Agents deployment.
Three Migration Strategies
1. Lift and Shift (Minimal Downtime)
Use the AgentCore Import-Agent toolkit. Converts your existing agent definition to LangGraph. Test in a staging environment with AgentCore Runtime. Flip traffic after validation. Timeline: 2-4 weeks.
2. Rebuild on AgentCore Native (Best Long-Term)
Rewrite your agent logic to use AgentCore Gateway for tool routing and AgentCore Memory for state. This gives you multi-agent orchestration from day one. Timeline: 4-8 weeks depending on complexity.
3. Hybrid (Lowest Risk)
Keep Bedrock Agents Classic running in maintenance mode. Build new features on AgentCore. Migrate incrementally as Classic reaches end of support. No hard deadline announced yet, but AWS typically gives 12 months of maintenance before full deprecation.
How ServerGurus Helps
We are an AWS Partner Network member. Our managed AWS team handles Bedrock migrations, cost optimization, and architecture design. If you have Bedrock Agents in production, we will audit your setup, build a migration timeline, and execute it without downtime.
If you are starting fresh with AI agents on AWS, we will architect your AgentCore deployment to avoid the cost traps and scaling issues that come with multi-agent orchestration.
Talk to our AWS team. Migration planning is free.