Harness is introducing the Agent-Ready Code Repository and AI Code Review. Both are designed to handle the deluge of code AI agents produce. Agents are assigned their own permissions, and pull requests are grouped by risk rather than by file.
According to Harness, indexing, search functions, and file history fall behind as volume increases. The permissions model is also problematic, as it’s based on a list of human developers, not on an agent that can merge code independently.
“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,” says Jyoti Bansal, CEO and co-founder of Harness.
Agents with their own permissions, review at risk
The Agent-Ready Harness Code Repository has been tested with thousands of concurrently open pull requests and commits, across repositories with tens of thousands of branches. Agents inherit permissions from the human who invokes them, after which developers can further restrict what an agent can access, merge, or deploy. RBAC and Open Policy Agent establish those limits before the merge.
Both the repository and the review are accessible via the Harness MCP and CLI. You can get started for free with 50 GB of storage. According to Harness, migrating from GitHub, GitLab, Bitbucket, or Azure DevOps takes just a few clicks, including pull requests, labels, webhooks, and branch rules.
AI Code Review groups diffs by risk rather than by file. Teams decide for themselves which AI Checks are mandatory. A change that fails such a check cannot be merged. Harness says this saves its own engineering teams more than 10,000 hours of manual review work each month.
Both features are available immediately.
Tip: Harness brings the entire agent lifecycle to a single platform