AWS supercharges AI automation with persistent browser agents

AWS has introduced Bedrock AgentCore Runtime Instances, a significant leap forward in enabling persistent multi-agent AI workflows on EC2 with session durations extending up to 14 days. This development marks a clear advancement over the previous microVM-based sessions, which were limited in duration and scope, allowing for more complex, stateful AI operations within managed cloud environments. By extending session longevity, AWS empowers enterprises to orchestrate sustained, coordinated AI agent activities that can handle intricate tasks without interruption, enhancing automation capabilities in legacy and modern applications alike.

Amazon Bedrock AgentCore Browser Tool revolutionizes automation of legacy web applications by enabling AI agents to drive secure, isolated browser sessions in the cloud through a managed Chromium instance integrated with Playwright and WebSocket-based Chrome DevTools Protocol connections. This approach overcomes the inherent limitations of traditional RPA, which often falters when navigating multi-step workflows, dynamic UI elements, and session-dependent state management, thereby allowing seamless interaction with legacy systems regardless of their underlying technology stack.

Legacy applications lacking modern APIs pose significant challenges due to complex authentication mechanisms such as multi-factor authentication, proprietary single sign-on, and unpredictable session token expiration, which traditional rules-based RPA bots struggle to handle reliably. The Browser Tool addresses these hurdles by automating interactions with HTML rendered by server-side middleware, drastically reducing costly manual data entry errors and enabling enterprises to modernize critical workflows that were previously resistant to automation.

By integrating with Strands Agents and Amazon Bedrock foundation models, the Browser Tool supports sophisticated, scalable automation workflows that range from single-step tasks to complex multi-agent orchestrations, all while maintaining session-isolated security enforced by AWS Identity and Access Management controls. This model-driven orchestration not only enhances operational efficiency but also ensures comprehensive audit trails and compliance with stringent regulations such as GDPR, HIPAA, and financial industry mandates, preserving essential human oversight in regulated sectors like healthcare, finance, and insurance.

Amazon Bedrock AgentCore revolutionizes enterprise AI orchestration by enabling sophisticated multi-agent collaboration within secure, managed EC2 environments. This platform offers fine-grained control over agent behavior, memory, and coordination, allowing organizations—such as M&A teams employing proprietary methodologies—to automate complex workflows with built-in governance, compliance, and auditability. Each agent action is traceable through citation-validated assertions and comprehensive audit trails, ensuring that AI-generated insights remain accurate, transparent, and policy-compliant throughout the workflow.

Bedrock AgentCore’s persistent memory and intelligent routing capabilities empower agents to build and share institutional knowledge over time, dramatically accelerating complex, multi-source data gathering and analysis. By enriching a shared memory layer with prior research, valuation assumptions, and integration lessons, agents efficiently prioritize high-value tasks and route opportunities to the right specialists, enabling enterprise-scale automation that transcends traditional task-based RPA limitations.

Security and compliance are deeply embedded in Bedrock AgentCore’s architecture through propagation of user authorization context and workload access tokens that bind user and agent identities. This design enforces least privilege access via infrastructure-level controls rather than agent code, employing session-tagged credentials and on-behalf-of token exchanges with services like Salesforce. Consequently, even if an agent is compromised, unauthorized data access is prevented, safeguarding sensitive workflows and supporting rigorous governance demands in regulated environments.

The introduction of the Bedrock AgentCore Gateway centralizes governance for AI agent tool access by providing a secure entry point that enforces authentication, authorization, and policy controls at scale. This gateway addresses enterprise challenges such as credential sprawl, policy drift, audit gaps, and shadow IT by cataloging tools, controlling access, auditing interactions, and hardening security at the edge. By layering these governance scopes, organizations achieve robust, compliant multi-agent coordination essential for automating complex workflows with confidence.