title: "AWS ships AgentCore runtime for 14-day agent sessions" slug: "aws-ships-agentcore-runtime-for-14-day-agent-sessions" published: "2026-10-01" beat: "Launches" tags: ["Launches", "Tools"] creator: "Agentry Newsroom" editor: "Susanne Sperling, Editor — Human in the Loop" tools: ["Claude (Anthropic)", "Perplexity Sonar"] creativeWorkStatus: "verified" dateReviewed: "2026-10-01" aiActArticle50: "compliant" humanView: "https://agentry.news/launches/aws-ships-agentcore-runtime-for-14-day-agent-sessions" agentView: "https://agentry.news/agent/aws-ships-agentcore-runtime-for-14-day-agent-sessions"
Amazon Web Services announced on September 9, 2026, that AgentCore Runtime instances can execute agents on dedicated Amazon EC2 compute for sessions lasting up to 14 days, and that AgentCore Payments
Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. AI policy.
Amazon Web Services announced two concrete shipping features for its Bedrock AgentCore platform on September 9, 2026: AgentCore Runtime instances that can run agents on dedicated compute for up to 14 days, and AgentCore Payments with infrastructure-enforced spending limits and end-to-end observability AWS.
The long-running runtime addresses a core constraint in agent deployment: traditional cloud functions and serverless compute have session time limits that force agent workloads into fragmented, stateless execution patterns. AWS said "AgentCore Runtime instances let you run agents on dedicated Amazon EC2 compute, including GPU-accelerated, memory-optimized, and compute-optimized instances, with sessions lasting up to 14 days" AWS. This removes the restart penalty for long-horizon tasks — research, analysis, or iterative problem-solving that cannot complete in minutes.
The second launch targets a governance problem: agents that can call external APIs and purchase content need hard limits on spend, not just logging. AWS stated that "AgentCore payments lets agents access and pay for APIs, MCP resources, and paid content with infrastructure-enforced spending limits and end-to-end observability" AWS. The emphasis on "infrastructure-enforced" means limits are applied at the platform layer, not by the agent itself — a critical distinction for preventing runaway costs or unauthorized expenditure.
Both features ship as generally available capabilities within Amazon Bedrock, AWS's managed AI service. Developers can access AgentCore Runtime through standard EC2 provisioning and AgentCore Payments through Bedrock's API, with observability integrated into CloudWatch and AWS's broader telemetry stack.
These are not roadmap announcements or capability claims. Both features are available for immediate use and directly remove two friction points in production agent deployment: session length and financial guardrails. For enterprises running agents on real business tasks — expense management, customer support, content analysis — the 14-day runtime eliminates the need to architect agent state externally or design for forced checkpoints, while the payment controls satisfy audit and compliance requirements that have blocked agent adoption in regulated industries.
The announcement reflects a shift in the agent infrastructure layer: from chatbot-style synchronous interactions to deployed, long-running autonomous systems that manage resources and external services.