Metrics Glossary
Definitions for all metrics across Intentra analytics and compliance pages.
MCP & Plugins
Calls Total MCP tool invocations in the selected time period.
Tools Number of distinct tools exposed by an MCP server.
Cost Estimated cost attributed to a server or plugin, calculated proportionally based on scan duration or call count.
$/Call Average cost per tool invocation. Calculated as total cost divided by call count.
Bloat Index Cost per call relative to the median across all servers. A value of 1.0x means average cost. Above 2.0x is flagged as bloated — the server costs significantly more per invocation than peers.
ROI Usage efficiency rating. High = frequently used with low error rates. Medium = moderate usage. Low = rarely used or error-prone.
Trust Score Security and reliability score (0–100) based on source verification (is it from a known publisher?), community health (GitHub activity), and usage reliability (error rates in your environment).
Risk Level Behavioral risk assessment combining trust score (40% weight) with behavioral signals (60% weight) like network access patterns, error bursts, and environment variable access.
Files
AI Sessions Number of AI coding sessions (individual tool invocations) that modified a given file.
AI % Percentage of code in a file attributed to AI generation, measured across snapshots over time.
Threshold AI content classification based on AI percentage: AI Dominant (>80%), AI Heavy (50–80%), AI Assisted (20–50%), Human Primary (<20%).
Drift Trend direction of AI content percentage over time. Increasing AI means more code is being AI-generated. Drifting to human means developers are replacing AI code.
Snapshots Number of times a file's AI attribution was measured. More snapshots = more reliable trend data.
Users
Cost Total estimated AI tool spend for a user in the selected period.
Scans Number of AI coding sessions analyzed. Each scan represents one tool invocation (e.g., one Cursor session, one Claude Code interaction).
Retry Rate Percentage of sessions that contained retry or recovery patterns, where the AI tool had to redo work. High retry rates indicate tool reliability issues.
Repositories
Retry Density Percentage of scans in a repository that contained retry patterns. Similar to Retry Rate but scoped to a specific repo.
Est. Cost/Branch Estimated cost per active branch, calculated as total repository cost divided by the number of branches with activity.
Stale Branches Branches with no AI coding activity for 30+ days. These may represent abandoned work with wasted AI tool spend.
Compliance
Overall Score Weighted compliance score from 0–100. 80–100: Passing. 60–79: Needs attention. Below 60: At risk. Based on the ratio of passing controls and severity of findings.
Last Assessed When the compliance engine last evaluated your scan data against enabled frameworks. Assessments run automatically when new data is available.
Controls Passing Number of compliance controls that currently meet all requirements with sufficient evidence, out of total evaluated controls.
Static Controls Controls that require manual verification (e.g., physical security policies, HR procedures) and cannot be assessed automatically. These are excluded from the automated score.
Evidence Coverage Percentage of controls that have automated evidence collection from your scan data. Higher coverage means more of your compliance posture is verified automatically.
Findings Issues identified during compliance assessment. Grouped by severity: Critical, High, Medium, Low. Address high-severity findings first.