Cursor Enterprise Analytics API: Setup Guide for Engineering Teams

Platform teams are asked to “meter Cursor” the same week finance notices the invoice doubled. The good news: Cursor exposes official read-only APIs for Enterprise teams. The bad news: those APIs are separate from billing, use different auth, and only cover part of what leaders want on a dashboard.

This guide walks through what to connect, what each endpoint gives you, and how to turn raw JSON into something a VP Engineering can act on — without proxies, browser extensions, or MITM tooling that security will block.

Who this is for

• Platform / developer experience engineers wiring internal analytics
• Engineering managers evaluating whether Cursor Enterprise includes the reporting they need
• FinOps or IT partners who need API-based exports instead of CSV archaeology

Prerequisites: Cursor Enterprise (Analytics API is not available on individual Pro plans). You need an admin who can create API keys in the Cursor dashboard.

Cursor gives you three layers of visibility

Most teams confuse these — they answer different questions:

Layer	Where	Best for
Web analytics	Cursor dashboard → Analytics	Quick charts, CSV export, 90-day filters
Analytics API	GET /analytics/team/*	Programmatic DAU, models, tabs, leaderboards
Admin / spend APIs	Team spend & daily usage endpoints	Per-user cost, billing groups, event-level detail

For a unified AI spend dashboard (Cursor + Copilot + OpenAI), you typically need Analytics API plus spend endpoints, then normalize into one schema alongside GitHub Copilot metrics and OpenAI org usage.

Step 1: Create an admin-scoped API key

1. Sign in to Cursor as an organization admin.
2. Open Dashboard → API Keys (or Settings → API Keys for your org).
3. Create a key with admin:* scope — required for Analytics API routes documented in Cursor’s Analytics API reference.
4. Store the key in your secrets manager (CURSOR_ADMIN_API_KEY). Never commit it to git.

Auth format: Basic authentication with the API key (see Cursor API overview for the exact header pattern your client expects).

Security posture to document for procurement:

• Read-only usage aggregation
• No access to source code or prompts in the analytics endpoints
• Key rotatable without developer IDE changes

Step 2: Know which endpoints to call first

Start with team-level routes that map to leadership questions:

Daily active users (DAU)

GET /analytics/team/dau

Answers: Is adoption growing week over week? A user counts as active if they use Tab, Agent, Background Agent, or CLI features (see Cursor docs for product-specific definitions).

Model usage

GET /analytics/team/models

Answers: Which models drive premium spend? Pair with spend APIs to see if claude-sonnet or gpt-4o defaults are bleeding budget on boilerplate tasks.

Tab vs agent mix

GET /analytics/team/tabs and agent-related endpoints

Answers: Are we paying for agent loops where tab completion would suffice? This is one of the fastest optimization levers — see agent vs tab completion tradeoffs.

Leaderboard

Team leaderboard endpoints (see current docs for path)

Answers: Who are champions vs who needs enablement? Use carefully — goal is coaching, not surveillance.

By-user pagination

Pattern: /analytics/by-user/{metric}

Use when you need per-developer reports or batch exports for >50 users. Supports pagination cursors for large orgs.

Step 3: Pull a 14-day baseline

Before alerting or optimizing, snapshot 14 contiguous days:

Day 0–13:  DAU, model mix, tab/agent counts
Parallel:  team spend by user (Admin API)

Normalize into one row per developer per day:

field	example
email	dev@company.com
provider	cursor
event_type	agent | tab | completion
model	claude-3.5-sonnet
tokens	124000
cost_usd	4.82
recorded_at	2026-06-28T00:00:00Z

ForgeMeter uses this shape so Cursor rows sit next to Copilot and OpenAI in the same Overview chart.

Common baseline mistakes:

• Comparing week 1 after rollout to week 12 — adoption curves distort trends
• Ignoring client version requirements (analytics collection needs recent Cursor desktop builds)
• Treating leaderboard rank as performance review input — kills honest usage

Step 4: Wire alerts to real thresholds

Once baseline exists, set thresholds finance will recognize:

Alert	Typical threshold	Owner
Daily spend spike	2× trailing 7-day average	Platform
Inactive seat / license	0 AI events in 30 days	IT
Model mix drift	>40% premium model tokens	Staff engineer
Team budget	80% of monthly allocation	Eng manager

Cursor’s native dashboard can filter users and export CSV — but it will not email your CFO when a runaway agent loop fires at 2 a.m. That’s where a metering layer (ForgeMeter, internal scripts, or cursor-usage-tracker) adds value.

Step 5: Combine with Copilot and API spend

Cursor rarely exists alone. In audits we see:

• Copilot seat still assigned while Cursor is primary IDE → double spend
• OpenAI project keys for CI while Cursor agents call the same models → invisible overlap

Minimum viable stack for engineering leadership:

1. Cursor Analytics + spend APIs
2. GitHub Copilot Usage Metrics API
3. OpenAI organization costs + completions usage (admin key)

Anthropic Admin API if Claude Code or API keys are in play.

API vs dashboard: when to build

Approach	Pros	Cons
Cursor web analytics only	Zero build	No cross-tool view, manual exports
Internal script + spreadsheet	Fast POC	Breaks on pagination / API changes
ForgeMeter / dedicated product	Unified schema, alerts, audit history	Another vendor to approve

Build internally if you have platform bandwidth and only Cursor to worry about.

Buy / adopt a meter when finance asks for one number across Cursor, Copilot, and OpenAI — or when you need weekly exec narratives without maintaining ETL.

Checklist: production-ready Cursor metering

• Enterprise contract confirmed (Analytics API enabled)
• Admin API key in secrets manager with rotation policy
• 14-day baseline stored with per-developer granularity
• DAU + model mix reviewed with eng leadership
• Spend spike alert tested against synthetic threshold
• Copilot + OpenAI connected or on roadmap
• Privacy / works council brief: usage metrics, not prompt content

How ForgeMeter helps

ForgeMeter connects your Cursor Admin API key alongside GitHub Copilot, OpenAI, and Anthropic. Sync now or daily auto-sync pulls official API data into one dashboard — per developer, per model, budget alerts, and email reports.

No traffic interception. Keys encrypted at rest.

Run a free AI engineering audit to baseline Cursor + Copilot + OpenAI in one report, or explore the live demo.

Related guides:

• How to track Cursor AI usage across your team
• GitHub Copilot cost analytics
• OpenAI Team Usage API for engineering