Cursor Usage AnalyticsSAFE

Model-agnostic agent-skills platform with a harness-free canonical layer, verified adapters, and the ccpi package manager. Explore at tonsofskills.com.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: cursor-usage-analytics
description: 'Track and analyze Cursor usage metrics via admin dashboard: requests,
  model usage, team productivity,

  and cost optimization. Triggers on "cursor analytics", "cursor usage", "cursor metrics",

  "cursor reporting", "cursor dashboard", "cursor ROI".

  '
allowed-tools: Read, Write, Edit, Bash(cmd:*)
version: 1.19.0
license: MIT
author: Jeremy Longshore <[email protected]>
tags:
- saas
- cursor
- analytics
- dashboard
compatibility: Designed for Claude Code
---
# Cursor Usage Analytics

## Overview

Use tenant analytics to improve enablement, capacity, and policy compliance with aggregated, purpose-limited data—not to inspect individual prompts or create unapproved employee surveillance.

## Prerequisites

- An approved analytics purpose, access role, retention rule, and privacy/HR review where applicable.
- Aggregated reports and a named owner for any action taken from the data.

## Instructions

1. Define the decision the metric supports and collect the minimum aggregation needed.
2. Review adoption, spend, model use, and policy signals against an agreed baseline.
3. Share only authorized aggregate findings and document remediation/enablement actions.
4. Reassess access and retention regularly; do not repurpose analytics without approval.

## Output

- A purpose-bound aggregate report with owner, time range, access boundary, and action items.

## Error Handling

| Condition | Safe response |
|---|---|
| Report exposes individual sensitive behavior | Restrict access, aggregate/redact further, and consult privacy/HR owners. |
| Metric is incomplete | Mark it as directional; do not make performance or spending decisions as if it were complete. |
| Spend anomaly appears | Verify authorized usage and tighten approved limits rather than inspecting prompt content. |

## Examples

Review monthly aggregate model spend by team against budget, publish the variance and owner, and adjust approved caps after review. Do not export individual chat content or infer productivity from request counts.

Track and analyze Cursor usage metrics for Business and Enterprise plans. Covers dashboard metrics, cost optimization, adoption tracking, and ROI measurement.

## Admin Dashboard Overview

Access: [cursor.com/settings](https://cursor.com/settings) > Team > Usage (Business/Enterprise only)

```
┌─ Dashboard ────────────────────────────────────────────┐
│                                                        │
│  Total Requests This Month:  12,847                    │
│  Fast Requests Remaining:    2,153 / 15,000            │
│  Active Users:               28 / 30 seats             │
│  Most Used Model:            Claude Sonnet (62%)       │
│                                                        │
│  ┌─ Usage Trend ─────────────────────────────────┐     │
│  │  ▆▆▇▇██▇▆▇█▇▇▆▅                              │     │
│  │  Mon Tue Wed Thu Fri Sat Sun                   │     │
│  └───────────────────────────────────────────────┘     │
│                                                        │
│  ┌─ Top Users ───────────────────────────────────┐     │
│  │  [email protected]     847 requests  (Sonnet)      │     │
│  │  [email protected]       623 requests  (GPT-4o)      │     │
│  │  [email protected]     591 requests  (Auto)        │     │
│  └───────────────────────────────────────────────┘     │
│                                                        │
└────────────────────────────────────────────────────────┘
```

## Key Metrics

### Request Metrics

| Metric | What It Measures | Target |
|--------|-----------------|--------|
| Total requests | All AI interactions (Chat, Composer, Inline Edit) | Growing month-over-month |
| Fast requests | Premium model uses (count against quota) | Stay under monthly limit |
| Slow requests | Queued requests after quota exceeded | Minimize (upgrade if high) |
| Tab acceptances | How often Tab suggestions are accepted | 30-50% acceptance rate is healthy |

### User Adoption Metrics

| Metric | Healthy | Needs Attention |
|--------|---------|-----------------|
| Weekly active users | 80%+ of seats | Below 50% of seats |
| Requests per user/day | 5-20 | Below 3 (underutilization) |
| Users with 0 requests (30d) | 0-10% of seats | Above 20% (wasted seats) |
| Model diversity | 2-3 models used | Single model only |

### Cost Metrics

| Metric | Calculation |
|--------|-------------|
| Cost per seat | Plan price / active users |
| Cost per request | Total spend / total requests |
| BYOK costs | Sum of API provider invoices |
| Total AI spend | Cursor subscription + BYOK costs |

## Quota Management

### Fast Request Quota

Each team member gets ~500 fast requests per month (varies by plan). Fast requests are consumed when using premium models (Claude Sonnet/Opus, GPT-4o, o1, etc.).

When quota is exceeded:

- Requests are queued as "slow" (may take 30-60 seconds instead of 5-10)
- Tab completion is unaffected
- cursor-small model remains fast

### Strategies to Stay Under Quota

```
1. Default to Auto mode
   - Cursor routes simple queries to cheaper models
   - Only uses premium models when complexity warrants it

2. Educate team on model selection
   - Simple questions → cursor-small or GPT-4o-mini
   - Standard coding → GPT-4o or Claude Sonnet
   - Hard problems only → Claude Opus, o1 (these burn quota fast)

3. Reduce round-trips
   - Write detailed prompts (fewer back-and-forth turns)
   - Use @Files instead of @Codebase (less context = faster)
   - Start new chats instead of continuing stale ones

4. BYOK for power users
   - Heavy users can use their own API keys
   - Their requests don't count against team quota
```

## Reporting for Stakeholders

### Monthly Report Template

```markdown
# Cursor Usage Report - [Month Year]

## Summary
- Active users: X / Y seats (X% utilization)
- Total AI requests: X,XXX
- Fast request quota usage: XX%
- Monthly cost: $X,XXX

## Adoption Trends
- New users onboarded: X
- Users showing increased us

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.