Amazon Bedrock AgentCore increases default runtime quota limits
AgentCore now supports 5,000 concurrent agent sessions by default in primary US regions, letting teams scale to production without quota increase requests.
View original announcement →Visual Summary
graph TD
A{{AgentCore Runtime Quotas}}:::announced
B((AI Agent Developers)):::external
C([5000 Concurrent Sessions]):::feature
D([200 Interactions/sec]):::feature
E([25 New Sessions/sec]):::feature
F(Amazon Bedrock AgentCore):::compute
G(AWS Service Quotas):::compute
H([AgentCore Harness]):::feature
B ==>|"deploys agents"| F
F ==>|"enforces"| A
A -->|"us-east-1, us-west-2"| C
A -->|"all regions"| D
A -->|"all regions"| E
A -->|"backs"| H
B -.->|"requests increase"| G
G -.->|"adjusts"| A
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
Amazon Bedrock AgentCore has raised its default runtime quota limits, enabling customers to scale agent-based workloads more aggressively without requiring upfront quota increase requests. The new defaults support up to 5,000 active concurrent sessions in US East (N. Virginia) and US West (Oregon), and 2,500 in all other supported regions. All regions now uniformly support 200 agent interactions per second and 25 new sessions created per second out of the box.
How It Works
- Active concurrent sessions define how many agent sessions can be simultaneously running within an account; the new default is 5,000 in us-east-1 and us-west-2, and 2,500 in all other AgentCore-supported regions.
- Agent interactions per second (200 RPS) governs the throughput rate at which agents can process requests across all active sessions in a given region, now standardized across every supported region.
- New sessions per second (25 SPS) controls the rate at which fresh agent sessions can be instantiated, preventing burst-induced cold-start bottlenecks at scale.
- AgentCore Runtime underpins these quotas; the harness resource shares the same limits since each harness is backed by a managed Runtime instance provisioned by AgentCore.
- All quotas remain adjustable via the AWS Service Quotas console, meaning teams with workloads exceeding even these new defaults can request further increases through standard support channels.
- The quota increases apply automatically to existing and new accounts — no configuration changes or redeployment of agents is required to benefit from the higher limits.
Why It's Important
- Eliminates early scaling friction: Previously, teams hitting default limits had to file quota increase requests before going to production, adding lead time; higher defaults let teams launch and scale without that gate.
- Enables enterprise-grade concurrency out of the box: 5,000 concurrent sessions in primary US regions is sufficient for many large-scale customer-facing deployments without any customization.
- Supports high-throughput agentic pipelines: At 200 interactions/second, teams can build real-time, event-driven agent workflows (e.g., document processing, customer support automation) that were previously constrained by lower defaults.
- Reduces operational overhead: Fewer quota increase tickets and approval cycles mean engineering teams spend more time building agents and less time managing infrastructure limits.
- Signals platform maturity: Proactively raising defaults indicates AWS is hardening AgentCore's infrastructure to support production-scale deployments, increasing confidence for enterprise adoption.
How It's Different
- Higher defaults vs. competitors: Many agent hosting platforms require manual capacity provisioning or explicit scaling configuration; AgentCore's increased defaults provide production-ready capacity automatically.
- Uniform throughput quotas across regions: The 200 interactions/second and 25 sessions/second limits are now consistent across all supported regions, simplifying multi-region architecture planning compared to region-specific throughput tiers.
- Adjustable via Service Quotas console: Unlike hard limits on some platforms, all AgentCore runtime quotas are soft limits that can be raised through the standard AWS Service Quotas workflow without requiring AWS account team involvement for typical increases.
- Integrated with the full AgentCore platform: These quotas govern not just raw compute but the entire session lifecycle — memory, identity, gateway, observability — meaning the increased limits benefit the entire agent stack, not just execution.
- Framework-agnostic scaling: The quota increases apply regardless of whether agents are built with LangChain, OpenAI Agents SDK, Strands SDK, or custom frameworks, preserving flexibility while improving scale.
When to Prefer It
- High-concurrency customer-facing applications: When deploying AI agents that serve thousands of simultaneous end users (e.g., customer support bots, virtual assistants), the 5,000-session default in primary US regions removes the need for upfront quota planning.
- Real-time data processing pipelines: When agents must process high volumes of events or documents in parallel at 200+ interactions/second, the new throughput defaults support these workloads without throttling.
- Rapid production launches: When time-to-market is critical and teams cannot afford delays from quota increase approval cycles, the higher defaults allow immediate production deployment.
- Multi-region active-active deployments: When distributing agent workloads across multiple AWS regions for resilience, the standardized 2,500-session and 200 RPS defaults in non-primary regions provide a predictable baseline for capacity planning.
- Enterprise pilots scaling to production: When an internal pilot unexpectedly gains traction and needs to scale quickly, the increased defaults provide headroom without requiring infrastructure rearchitecting.
- Batch and scheduled agentic workflows: When running periodic high-volume agent tasks (e.g., nightly report generation, bulk data enrichment), the higher session and throughput limits reduce job completion time by enabling greater parallelism.
Availability
- General Availability: AgentCore Runtime and these quota increases are generally available; no preview or beta restrictions apply to the runtime quota changes.
- Primary US regions (5,000 concurrent sessions): US East (N. Virginia) and US West (Oregon) receive the highest default session limits.
- All other supported regions (2,500 concurrent sessions): Includes US East (Ohio), Europe (Frankfurt, Ireland, London, Milan, Paris, Spain, Stockholm), Asia Pacific (Malaysia, Mumbai, Singapore, Sydney, Thailand, Tokyo, Seoul), Canada (Central), South America (São Paulo), and AWS GovCloud (US-West).
- Throughput quotas (200 interactions/second, 25 sessions/second): Uniformly applied across all regions where AgentCore Runtime is available, including GovCloud.
- Adjustable limits: All runtime quotas are soft limits and can be increased via the AWS Service Quotas console; hard limits exist for some other AgentCore resources (e.g., versions per agent).
- Pricing: No additional charge for the quota increase itself; standard AgentCore Runtime pricing applies based on session duration and agent interactions.
- Automatic application: Existing accounts benefit from the new defaults immediately with no action required.
