Harness has launched Agent-Ready Harness Code Repository and AI Code Review, aimed at software teams using AI coding agents.
The new repository and review tools are being introduced as a linked system. Harness argues that code generated by AI agents is changing the volume and pace of software development, putting pressure on older source code management and review processes built around human developers opening and discussing pull requests over longer periods.
At the centre of the launch is a code repository designed for large spikes in commits and pull requests. Harness said the system has been tested to handle thousands of pull requests and commits opened at the same time while maintaining search, history and diff functions across repositories of different sizes and with very large numbers of branches.
Another part of the release focuses on access control for AI agents. Agents can inherit permissions from the human users who trigger them, while developers can narrow those permissions to particular repositories, branches or environments. The setup also supports role-based access control and Open Policy Agent policies to enforce boundaries before code is merged.
The tools are also available through the command line using the Harness MCP and CLI. Harness said this allows users to manage the full pull request process without opening a browser, including finding reviews by an author's email address, viewing open pull requests across repositories, and creating or resolving comment threads. AI agents can also use CLI commands to carry out repository and review actions in headless mode, according to the company.
Migration is another part of the pitch. Teams can start with a free tier that includes 50 GB of storage and move repositories from GitHub, GitLab, Bitbucket or Azure DevOps, with pull requests, labels, webhooks and branch rules transferred through the Harness CLI.
Review layer
The AI Code Review product is intended to reduce the burden on human reviewers as code volumes rise. Teams can decide which AI checks are mandatory and apply them at account or project level, with failed required checks blocking a merge.
The review process also changes how code differences are presented. Instead of grouping changes by file, the system groups them by risk, with the aim of highlighting changes that alter software behaviour ahead of lower-priority modifications such as file renames or dependency updates.
Harness said its review tool also offers what it describes as one-click remediation. Feedback is framed around the implications of a code change, while suggested reviewers and labels are added before a pull request is opened. If the feedback is accepted, changes can then be merged with a single click, the company said.
Jyoti Bansal, Chief Executive Officer and Co-Founder of Harness, set the launch in the context of wider changes in software engineering workflows.
"Software delivery is going through its biggest shift since the move to the cloud, and the systems we all built our workflows around were designed for a different scale and a different kind of user," said Jyoti Bansal, Chief Executive Officer and Co-Founder of Harness.
He added that the issue could not be addressed by adding AI functions to older repository products.
"You do not solve that by adding AI features to a repository designed fifteen years ago. The entire SDLC has to become autonomous, which means the repository, the review, the pipeline, and the governance must all work as one system," said Bansal.
Platform push
The launch also reflects Harness's broader effort to position its software delivery products as a single workflow from code commit through to production. Harness said committing, reviewing, building, testing, securing and deploying software already take place in one sequence inside its platform, and the new repository and review functions now sit at the start of that chain under one policy engine.
The review system draws on what Harness calls its SDLC Knowledge Graph, which maps release processes, policy requirements and production failures. According to the company, that information is used to provide context for code reviews.
Internally, Harness said its own engineering teams have been using the products for several months and that this has cut manual review work by more than 10,000 hours a month.
The products are available immediately as part of the company's wider push into AI-assisted software delivery.