evidra from Vitas is an AI-driven infrastructure flight recorder that captures agent intent, outcomes, and audit evidence for cloud change management. The tool records agent commands, summarizes command-line outputs, and generates reliability scorecards to evaluate infrastructure mutations across environments. It integrates as a Model Context Protocol server and a cross-platform CLI, with specialized adapters for Terraform and Kubernetes. DevOps engineers, SREs, and platform developers use it to add measurable accountability and traceability to AI-driven CI/CD workflows.

What tasks can you actually use it for?

The tool records and documents AI-driven infrastructure work, acting as a bridge between language-model clients and cloud environments. It captures the full context of agent interactions, including stated intent, executed commands, and final outcomes. Typical tasks include auditing agent-led deployments, producing condensed summaries of verbose CLI output, and collecting diagnostics after a failed change, which helps teams reconstruct what an agent attempted and why.

How accurate and actionable are the recorded outputs?

Recorded outputs are shaped for human review and evaluation; the system generates Reliability Scorecards that evaluate success and safety, and uses AI summarization to condense complex terminal output into concise insights. The tool also auto-collects logs and diagnostics as evidence for each mutation, which turns raw command output and artifacts into consumable material for incident reviews and postmortems.

What file formats and integrations does it accept?

The tool fits standard infrastructure toolchains by integrating with the Model Context Protocol and common cloud providers. It supports interaction with AI clients such as Claude Desktop via MCP, and includes built-in tools for Kubernetes, Terraform, AWS, GCP, and Azure. Teams can expect adapters for kubectl and Terraform plan/apply workflows, and recorded artifacts that reference executed commands and captured logs.

Is it straightforward to add into existing DevOps workflows?

Deployment options match typical platform needs: the app runs as a standalone CLI or as an MCP server, so teams can host it locally or alongside existing control-plane services. Its design targets DevOps engineers and SREs, aiming to plug into CI/CD or platform automation pipelines where agent-driven changes require auditability and post-action evidence.

Practical choice for teams that need verifiable agent accountability

The tool is well received in MCP and AI-coding communities and suits DevOps engineers, SREs, and platform developers who require measurable audit trails for agent-driven changes. Organizations gain most when they pair the recorder with formal review and governance processes, because the tool produces examinable artifacts that require human workflows to convert findings into safer deployment practices.