Amazon Bedrock AgentCore harness is now generally available: Go from idea to production-grade agent in minutes

This article announces the general availability of Amazon Bedrock AgentCore harness, a managed service that simplifies deploying production-grade AI agents from idea to production in minutes.

• Deploy agents with two API calls (CreateHarness and InvokeHarness) instead of weeks of infrastructure setup
• Switch between multiple model providers mid-session while maintaining context (Bedrock, OpenAI, Gemini, LiteLLM)
• Declare tools as configuration without writing glue code: browser, code interpreter, MCP servers, and gateways
• Automatic managed memory with semantic and summarization strategies for multi-session conversations
• Built-in skills including AWS-curated catalog covering SDK usage, IaC, IAM, CloudWatch, and service-specific workflows
• Unified observability in CloudWatch with harness-level tracing across all primitives
• Evaluations and optimization features for scoring agent performance and A/B testing prompt changes
• Versioning and endpoints for safe rollouts with instant rollback capability
• Export to code (Strands-based) when configuration needs exceed requirements
• Pay-per-use pricing for runtime compute, browser, code interpreter, gateway, memory, and observability

AgentCore harness eliminates infrastructure overhead, enabling teams to rapidly prototype and deploy agents while maintaining production-grade reliability, observability, and continuous improvement capabilities.