Harness adds Cosmos to its software delivery platform

Harness says Cosmos is joining its platform alongside agents for software delivery, security testing, runtime protection and cost management.[2] Its announcement was dated October 8, 2026.[1] Augment says it agreed to sell select assets because it believes Harness’s integrated platform can deliver these capabilities more effectively than building them independently.[4] Augment co-founder and CTO Igor Ostrovsky described joining Harness as a step toward bringing AI to the entire software development lifecycle.[5]

Cosmos can start work from events such as ticket updates, pull requests, schedules and webhooks, rather than requiring someone to launch each task manually, according to Harness.[2] Harness says its agents can plan a change, write code and tests, open a pull request and respond to feedback.[2] Work can continue through review comments, failed checks and merge conflicts.[2] Harness says the work happens in a sandboxed virtual machine.[3] For teams assessing the workflow, the useful question is which recurring tasks are suitable for event-driven execution and where human checkpoints should remain.

Harness says teams decide what agents may do independently and where people must review or approve their work.[2] Teams can also adapt prebuilt Experts—including PR Author, Deep Reviewer, Risk Analyzer and Verifier—to their organizations’ needs.[2] Cosmos includes model routing, tool integrations, shared organizational memory, execution environments, and configuration and budget controls, according to Harness.[2] These options make governance and workflow design part of the adoption decision: teams can define approval boundaries and tailor agent roles rather than treating every task as fully autonomous.

Harness says the acquisition creates an opportunity to connect Augment’s Code Context Engine with Harness’s Software Delivery Knowledge Graph.[2] The graph connects context about how software is built, tested, secured, deployed and operated.[2] Harness says testing expectations, security requirements, deployment constraints and operational experience can inform engineering work before a change is written.[2] Findings from code review, quality and security testing, and deployment verification can also feed back into engineering work with evidence to investigate, correct and revalidate changes.[2] Harness frames the outcome as valuable software reaching users, rather than simply generating more code.[2]