Harness has acquired select assets from Augment Code to expand automation across the software development life cycle (SDLC). The acquisition includes Cosmos, the Auggie CLI products, the Code Context Engine, and related technology. The teams behind these products will join Harness as the company works to automate more stages of software development.
Cosmos will be introduced as the Harness Cosmos Software Factory Agent, responsible for taking engineering tasks from an initial idea to merge-ready code. Harness plans to connect it with its existing agents for software delivery, security testing, runtime protection, and cost management, linking code creation with the processes required to move software into production.
The announcement centres on connecting codebase knowledge with information from software delivery and production systems. This approach is intended to help AI agents make informed changes, identify problems, and address issues using feedback from later stages of development.
“AI’s impact on software will not be measured by how much code it generates, but by how much valuable software reaches customers,” said Jyoti Bansal, founder and CEO of Harness.
He added that the integration would allow downstream agents to send issues identified during testing, security checks, or deployment back to Cosmos, along with the context needed to fix them.
Harness Cosmos aims to automate engineering tasks
The Harness Cosmos Software Factory Agent is intended to handle tasks triggered by written requirements, assigned tickets, and reported bugs. A fleet of agents plans changes, writes code and tests, and opens pull requests for review.
Each agent operates in an isolated virtual machine (VM). Engineers are involved when human judgment is required, such as approving a design or completing the final merge.
The system follows a continuous workflow covering tickets, code, and reviews. If a reviewer raises a concern or an automated check fails, an agent can make corrections on the same pull request.
Teams can also adapt prebuilt agents called Project Builder, PR Author, Deep Reviewer, and PR Fixer to their codebases and deploy them across their organisations.
The Code Context Engine maintains a live map of the codebase to help agents account for existing code and dependencies. Model routing assigns tasks to appropriate artificial intelligence (AI) models, while integrations with GitHub, Jira, and Slack connect the workflow with commonly used engineering tools.
Shared memory allows the system to carry lessons from earlier reviews into subsequent changes. Versioning and budget controls are also part of the offering.
Connecting code context with delivery data
The acquisition brings Augment Code’s Code Context Engine together with Harness’s Software Delivery Knowledge Graph. The latter connects information about builds, deployments, infrastructure, security, and costs.
Harness says Cosmos will use delivery information before making changes, including relevant test cases, security requirements, previous incidents, and their resolutions. Once code is created, downstream agents can use codebase information to assess dependencies, risks, and required validation.
If testing or deployment identifies a problem, the agents can return the issue to the coding workflow for correction and verification.
Igor Ostrovsky, co-founder and CTO of Augment Code, said joining Harness was a step toward the company’s goal of bringing AI across the software development life cycle, beginning with context-aware coding.
He said the combined effort could make more engineering work autonomous while allowing engineers to focus on decisions requiring human judgment.
Extending automation from development to production
Harness already offers agents covering several stages beyond code creation. Its Software Delivery Agent handles the movement of changes into production, while the Security Testing Agent addresses risks before release. The Runtime Protection Agent focuses on applications, application programming interfaces (APIs), and AI in production. The Cost Management Agent manages cloud and AI spending.
The integration plan also covers policies, permissions, and approvals already used across the Harness platform. The stated aim is to apply the same controls to code creation that enterprises use when releasing software.
The announcement outlines a broader approach to autonomous software development in which AI agents handle repetitive engineering tasks and exchange information across the development pipeline. Engineers would retain responsibility for decisions requiring judgment, while feedback from testing, security, and production could guide subsequent code changes.
The announcement did not specify the financial terms of the acquisition or provide a timeline for the availability of the integrated offering.