Help AI-built software meet your engineering standards

Assurance Agents bring your standards into implementation, then independently review and test the software against the applicable requirements and standards. Teams can see what needs correction and what remains unresolved before release.

Apply your standards without manually repeating every check

Coding agents can produce changes faster than experts can review them. Generic checks may miss the architecture constraints, security requirements, or compatibility needs specific to your software. Senior engineers then have to find and explain the same violations across changes.

Assurance Agents select the standards relevant to the repository and task so coding agents can apply them while they work. They then independently check the implementation against those standards and the applicable specification, using review and focused tests to identify gaps.

Review and test changes during development, before merge, and in CI

Apply assurance during development, before merge, and in CI. Each evaluation uses the requirements, standards, and evidence relevant to its scope.

  • During development

    Review the implementation and test relevant behaviour. Return findings that developers or coding agents can use to revise the change.

  • Before merge

    Evaluate the proposed change with the relevant repository files and supporting artifacts. Identify gaps that need correction, additional tests, or a reviewer decision.

  • In CI

    Audit the broader repository against agreed standards and apply the review or release policy chosen by the engineering team.

Give engineers the evidence to correct and reassess a change

Findings connect an issue to the applicable requirement or standard, affected code or configuration, supporting evidence, and severity. They explain why it matters and what correction or further verification is needed.

Developers or coding agents can revise the implementation and run the checks again. Reviewers can see what was checked and what remains unresolved. The engineering team owns accepted exceptions and decides which findings advise, require human review, or block progression. It retains the release decision.

Start with Kubernetes Assurance

Kubernetes Assurance is the first agent being developed as a proof of concept. It applies your cloud native and security standards to application code, Kubernetes configuration, and rendered deployment output during development, pull request review, and repository audits in CI.

Extend assurance to security, contracts, and upgrades

Ajwain is exploring additional assurance areas where teams have explicit standards and difficult review decisions. These are proposed extensions to the current Kubernetes work, with scope to be defined and evaluated against real changes.

Start with a change that requires expert review

Bring your repositories, Kubernetes changes that require expert review, and the engineering standards that apply. Agree a focused proof of concept with the reviewers who own those decisions. Assess the findings, missed issues, false positives, and usefulness of the proposed corrections before deciding which checks to enforce.

Discuss an assurance proof of concept