Harness has acquired Augment Code’s Cosmos software factory, including its AI-powered coding agents and context engine, aiming to embed autonomous coding deeply into the software delivery lifecycle and improve release reliability through smarter, context-aware automation.
- Introduces autonomous coding agents connected to deployment and testing pipelines.
- Enables iterative code improvement with feedback from previous failures.
- Supports customizable agent workflows integrated with Harness’s delivery graph.
Infrastructure signal
Harness’s acquisition of Augment Code’s Cosmos software factory marks a key evolution in cloud-native infrastructure by incorporating AI-powered autonomous coding agents directly into the software delivery platform. These agents operate within isolated virtual machines, generating and testing code while accessing deep codebase context and deployment history.
By connecting Cosmos with Harness’s Software Delivery Knowledge Graph, the platform can use real-time feedback from downstream testing, security, and deployment processes to direct coding agents toward necessary fixes automatically. This feedback loop reduces manual intervention and has the potential to lower cloud operational costs by accelerating the release pipeline and decreasing failure impact.
Developer impact
This approach aims to alleviate the AI velocity paradox, where rapid code generation overwhelms downstream testing and release systems, by enabling coding agents to incorporate failed test results and earlier review comments into subsequent iterations. This iterative correction model promises improved code quality and reduced rework in development cycles.
What teams should watch
Teams should monitor how Harness rolls out integration between Cosmos’s Code Context Engine and its delivery graph, particularly observing how deployment and security feedback is shared with coding agents. Understanding the mechanisms for permission and identity management across these AI-driven components will also be critical as agent identity models mature industry-wide.
Additionally, teams engaged in platform engineering, release management, and developer productivity should evaluate how much autonomy to grant AI agents within their workflows. The flexibility to fork and customize these agents presents opportunities but requires governance to balance speed, reliability, and security in continuous delivery environments.