AWS previews Claude Mythos on Bedrock
AWS has made Claude Mythos available in a gated preview through Bedrock, offering early access to allowlisted organisations under a project named Glasswing, and also introduced an Agent Registry for managing AI agents and tools. The preview appears targeted at advanced security and code‑analysis use cases. (x.com) (x.com)
Amazon Web Services has opened a gated Bedrock preview of Anthropic’s Claude Mythos, with access limited to invited organisations under Project Glasswing. (aws.amazon.com) Amazon posted the Bedrock launch on April 7, 2026, and its documentation says Claude Mythos Preview is aimed at cybersecurity, autonomous coding, and long-running agents. The same page says access is being prioritized for defensive cybersecurity use cases. (aws.amazon.com) (docs.aws.amazon.com) Bedrock is Amazon’s managed service for running foundation models from companies including Anthropic, and the Mythos listing places this release inside Amazon’s existing model marketplace rather than as a standalone Anthropic product. AWS says customers must be allowlisted for the research preview. (aws.amazon.com) (docs.aws.amazon.com) The setup points to a narrower release than a normal model launch. Anthropic’s model card, as published in AWS documentation, says Mythos is built for “ambitious projects” involving software security reviews, coding, and agents that run for extended periods with less human intervention. (docs.aws.amazon.com) AWS paired that preview with a second release on April 9: Agent Registry, a private catalog inside Amazon Bedrock AgentCore for agents, tools, skills, Model Context Protocol servers, and custom resources. AWS says the registry includes approval workflows, semantic and keyword search, and Amazon CloudTrail audit logs. (aws.amazon.com) (docs.aws.amazon.com) AgentCore is Amazon’s managed layer for deploying and operating artificial intelligence agents, and the registry gives companies a way to track which agents and tools already exist before teams build duplicates. AWS says both human users and agents can search the registry for approved resources. (docs.aws.amazon.com 1) (docs.aws.amazon.com 2) Taken together, the two launches show Amazon pushing Bedrock beyond chatbots and into governed agent operations. One release adds a restricted model for security and code analysis; the other adds inventory and controls for the agents and tools that might use it. (aws.amazon.com 1) (aws.amazon.com 2) Amazon has not announced general availability, pricing, or a broad customer rollout for Claude Mythos on Bedrock. For now, AWS is treating Mythos as a research preview and Agent Registry as a preview, which keeps both products inside controlled early access while Amazon tests how far customers want to automate security and software work. (aws.amazon.com 1) (aws.amazon.com 2)
