Rubrik has launched Rubrik Agent Cloud for Anthropic's Claude Code and Claude Cowork, aimed at organisations deploying Claude-based agents in software development workflows.
The service is designed to add oversight, recovery and policy controls around autonomous agents that can write, push and deploy code without direct human action. Rubrik positions the launch as a response to a security gap between traditional enterprise controls and AI systems that can act at machine speed.
The new service includes what Rubrik calls a governance engine for real-time control of autonomous agents, alongside an inventory tool intended to show risk, access permissions and policy violations across deployed agents. It also includes a feature called Agent Rewind, intended to reverse unintended actions taken by custom agents and agents built in tools such as Claude Code and Claude Cowork.
Another part of the launch focuses on code repositories. Rubrik says it maintains immutable snapshots of GitHub and Azure DevOps repositories outside the repository environment, allowing organisations to restore a known-good state if version control cannot undo an unwanted change, such as deleted branches or overwritten commit history.
The company is also offering backup and restoration for the configuration that shapes how Claude agents behave. That includes system prompts, tool permissions, skills and key files used at organisation, repository and user levels. The service monitors for configuration drift and flags changes that appear malicious or unauthorised before they spread.
The move reflects a broader shift in software development, where AI agents are taking on a larger role in producing and modifying code. That has created new concerns for security and compliance teams because existing DevSecOps processes were largely built around human approvals and manual review.
Rubrik identified risks including rogue commits, repository ransomware, prompt injection and intellectual property exfiltration. In its view, those threats become harder to contain when an automated agent can make changes quickly and at scale.
Anneka Gupta, Chief Product Officer at Rubrik, commented on the demand the company is seeing around Anthropic's models.
"Organisations are adopting Claude faster than any agentic technology we have seen, and every security leader asks the same question: how do we stay in control when an agent can act?" said Anneka Gupta, Chief Product Officer at Rubrik.
She said the product is intended to give customers a way to govern access and recover from unwanted actions affecting both code and agent settings.
"Rubrik Agent Cloud gives organisations a resilience layer for Claude, which allows them to see what agents can access, govern what they do, rewind their actions, and recover both the code and the agent's own configuration when something is destroyed or tampered with. Working with Anthropic, a leader in AI, lets us bring that control to customers from day one," Gupta said.
Security focus
The launch underlines an emerging market for AI operational controls, with suppliers trying to address governance, auditability and rollback for autonomous software agents. As businesses experiment with AI tools that can interact directly with codebases and internal systems, security teams are under pressure to track what those systems can access and what they have changed.
Rubrik's approach combines monitoring with recovery. If an agent, or an attacker exploiting one, takes an action outside the reach of normal version control, administrators can restore repositories at either repository or organisation level. Rubrik also says ransomware rollback is part of the offer for code environments.
Anthropic link
The launch centres on Anthropic's Claude tools, which are increasingly being used in coding and workplace assistant tasks. Anthropic describes itself as an AI safety and research company and says the Claude family of models is used for work including code understanding and security analysis.
For Rubrik, the tie-up expands its push into AI operations alongside its established work in cyber resilience and data protection. The company has been building products aimed at helping customers monitor, audit and recover AI-driven environments as the use of autonomous agents moves from experimentation to production.
The central issue for customers is likely to be whether safeguards can keep pace with the increasing autonomy of AI systems. In software development settings, that means not only preventing unauthorised or damaging changes but also restoring systems quickly when prevention fails.
Rubrik says its service tracks and restores the files and permissions that guide agent behaviour, including settings such as CLAUDE.md, system prompts and tool access. It also ties detected configuration drift to healthy backup snapshots to support orchestrated recovery.
That emphasis on both code and configuration reflects growing recognition that AI agents introduce a second layer of operational risk. The threat is not limited to what an agent changes in a repository, but also extends to whether the rules governing that agent have themselves been altered without approval.
Rubrik says Agent Rewind can reverse unintended actions across custom agents and common agentic development environments, including Claude Code and Claude Cowork.