Anthropic is building some amazing stuff, and as a Claude user myself, I always look forward to trying it out, Claude Managed Agents included. Some may wonder, how does Claude Managed Agents compare to GitLab's Duo Agent Platform (DAP), because on the surface they may sound very similar? From my time studying it, I think of Claude Managed Agents as a super powerful agent harness that runs in their cloud with all the basic tools (bash, file ops, web search). It a very powerful building "engine" that can be applied to many knowledge worker tasks. It is not a software factory, for example it has no first class concept of a repo, a merge request, a pipeline, or a security policy. DAP is a software factory, with all those primitives built-in. Most importantly, it runs within the organization's workflows and guardrails, and provides access to unified SDLC context. DAP agents have first class access to the entire repo, not just local code, they have access to pipeline state, MR history, security findings and much more. DAP provides the organization full governance and control: verification of the code quality and that it meets their engineering standards, that agents are operating within existing permissions and guardrails, with full traceability. DAP provides vulnerability scanning, vulnerability management, policy enforcement as first-class inputs. DAP is cloud neutral, runs everywhere including your datacenter or air gapped environment and is model-agnostic. DAP supports Claude code, and Codex agents built-in, but also supports Duo agents or you own custom agents. In summary, Claude Managed Agents is a really powerful Anthropic-only "engine" that could power many knowledge worker tasks, including software coding. GitLab's Duo Agent Platform is the "factory floor" that orchestrates engines - including Claude - and governs what they produce, with integrated context across the full software development lifecycle.