Agent Pro+ usage guidance contradicts docs and live dashboard

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Agent product knowledge for Pro+ usage monitoring does not align with official Cursor docs or instructions, or the live dashboard UI.

User on Pro+ asked how to monitor usage and how long agentic work can be about one tenth Claude cost.

Agent answered with meter semantics (watch Spending percent, ignore Usage dollars) and opened Overview. Overview only shows On-Demand Usage this Month of 0 of 5 dollars plus Edit Limit. Included API and First-party pools are under separate sidebar items named Usage and Spending.

Official docs describe two pools: First-party (Auto, Composer, Grok) and API (frontier models). Pro Plus includes about 70 dollars of API usage plus a generous first-party pool.

Agent session strip showed plan as pro, limit n slash a, and login token expired while the dashboard showed Pro Plus.

Also the agent skill path under .cursor slash skills-cursor slash cursor-guide slash SKILL.md is referenced for product Q and A but missing on disk, so the Agent falls back to stale knowledge.

Related: forum topic 166052 about open_resource Chrome Beta vs agent Browser Tab login confusion.

Steps to Reproduce

  1. On a Pro Plus account, ask Agent how to monitor usage and how agentic work is cheaper than Claude.
  2. Compare the Agent answer to official Models and Pricing docs and to the account dashboard Overview page.
  3. Note that Overview shows only On-Demand 0 of 5 dollars.
  4. Note Agent session strip plan label versus dashboard Pro Plus.

Expected Behavior

Agent product Q and A about billing must be grounded in current docs and the live dashboard. Direct users to Usage and Spending, not Overview-only On-Demand. Describe First-party versus API pools correctly for Pro Plus. Use live account plan Pro Plus, not stale pro or expired token context. Product guidance skill cursor-guide should exist and be used.

Operating System

MacOS

Version Information

Version: 3.12.17
Commit: 0fb762053c34788bb7760d5673f8a6d4c8589d50
OS: Darwin arm64 25.5.0

For AI issues: which model did you use?

Composer / Agent

Additional Information

Screenshot evidence: dashboard Overview with Pro Plus and On-Demand 0 of 5. User said Agent knowledge does not align with instructions. Zero URLs in this post due to new-user link limit.

Does this stop you from using Cursor

No - Cursor works, but with this issue

Your reading is the correct one. The docs and your dashboard are right, and the Agent’s billing answer was off. A few clarifications:

Where to monitor usage: check the Usage section at cursor.com/dashboard. That’s where your real-time usage, remaining allowance, and token breakdowns live. The Overview page’s “On-Demand Usage this month” ($0 of $5) is only the pay-as-you-go overage bucket, not your included usage, so it’s expected to stay at $0 until you go over your plan.

The two pools (Pro+): you’ve got it right. There’s a First-party models pool (Auto, Composer 2.5, Grok 4.5) with a generous included allowance, plus an API pool ($70 included on Pro+) for frontier models. First-party models draw from that more generous pool, which is why leaning on Auto/Composer stretches much further than picking a specific frontier model. Both are laid out here: Models & Pricing and Usage and limits.

On the Agent’s answer: its billing/product replies are model-generated and can lag product changes, so treat the dashboard and docs as the source of truth for usage questions. Asking the Agent to check the Cursor docs on a product question tends to give you a grounded answer.

Two smaller things you flagged: the .cursor/skills-cursor/cursor-guide/SKILL.md path was the Agent inventing a detail, not a real Cursor file, so nothing is missing or broken on your machine. And the “plan: pro / token expired” note in the chat context isn’t reliable account state. Your dashboard is correct that you’re on Pro+.

Let me know if anything in the Usage view still doesn’t line up!