SecurityBrief India - Technology news for CISOs & cybersecurity decision-makers
India
Reva.AI launches security tool for coding agents

Reva.AI launches security tool for coding agents

Thu, 19th Nov 2026 (Today)
Sofiah Nichole Salivio
SOFIAH NICHOLE SALIVIO News Editor

Reva.AI has launched Coding Agent Security, a product for AI coding agents, with a self-serve 15-day trial.

The launch addresses a growing concern for software and security teams as AI coding agents take actions on behalf of developers, including editing files, running shell commands and calling internal tools.

The product gives security and engineering leaders visibility into those actions, applies runtime authorisation to agent activity and records each action in an audit log. It is designed to work across Claude Code, Codex, GitHub Copilot and Kiro.

The system evaluates every action a coding agent attempts against policy and checks whether it aligns with the task assigned to the agent before execution. Teams can also start in monitoring mode before moving to enforcement.

It is designed to address the gap between an agent having technical permission to reach a system and whether a specific action is appropriate to the task it has been given. In practice, that can mean an agent with valid access still takes an action that affects the wrong branch, exposes a credential or changes records outside its intended scope.

Runtime checks

Reva.AI said its approach is based on what it calls Intent and Behaviour-Based Access Control, or IBAC. The model checks the purpose of an action alongside more conventional role-based and attribute-based policies.

Under the model, organisations can set central policies for coding agents and apply them across different tools. Customers can also require human approval for higher-risk actions.

That focus reflects a broader shift in how AI tools are used inside engineering teams. Coding assistants originally concentrated on generating snippets of code or suggestions inside an editor, but newer coding agents can perform sequences of actions with less direct supervision from developers.

As a result, the security question is shifting from what data an AI tool can see to what actions it can take once connected to repositories, command-line environments and internal services. Security teams also face the challenge of establishing a record of what an agent did, who initiated the task and why a particular action was allowed or blocked.

Reva.AI said its product is intended to create that record by tracing each agent action to the developer and task, and logging each decision for later investigation or audit.

Market context

Reva.AI describes itself as an authorisation management and AI runtime security business. It said the founding team includes identity specialists from Saviynt.

Independent industry recognition has begun to emerge around this segment. Reva.AI said analyst firm KuppingerCole named it a Rising Star in Authorisation Management and AI Runtime Security and gave it a 5 out of 5 score for innovation.

The launch comes as companies experiment with giving AI agents more autonomy in software development workflows while trying to avoid unintended changes or unauthorised access. For engineering leaders, that raises operational questions about how to let agents handle more routine or repetitive work without creating blind spots for audit and control.

For security teams, the challenge is not only prevention but also visibility. If an AI agent executes commands, opens or modifies files, or connects to internal systems, investigators may need to know which developer initiated the work, what task the agent was meant to complete and whether the action matched policy at the moment it occurred.

Reva.AI is positioning the product around those needs, with an emphasis on monitoring, policy checks and decision records rather than relying solely on the underlying permissions granted to the agent.

"Enterprises are already relying on coding agents for a growing share of their engineering work, and they need a central access control plane to govern those agents without slowing developers down," said Yash Prakash, co-founder and CSO of Reva.AI.

"Coding Agent Security checks every action against policy and the developer's intent before it runs, giving teams deep visibility and precision control over their agents' actions," Prakash said.