Pool 1 · metered
API Budget
Metered/premium/third-party usage draws from the API pool at the selected model’s API rate. On-demand pay-as-you-go and spend caps are explicit controls.
Cursor provider verdict · snapshot 2026-07-20
Is Cursor worth it if Auto’s exact cap is unpublished?
Currentness: Source audit is overdue or does not match this snapshot. This is a modeled decision aid, not a current recommendation.
Yes for IDE-first developers and teams who want integrated agent workflows and set spend caps. Avoid it when your buyer question requires a published Auto cap, strict task-level chargeback, or predictable CLI/API metering.
Official facts: Pro is $20/mo with $20 API agent usage; Pro Plus is $60/mo with $70; Ultra is $200/mo with $400. Cursor documents a separate Auto + Composer pool as “generous included usage,” but does not publish an exact Auto/Composer cap.
Lived reliability · Jul 12–18, 2026Insufficient current evidence
A supported access-interruption warning exists; current evidence is insufficient for cross-billing or broad quality-decline claims.
Decision: Treat invoice-state interruption as a workflow risk. Do not change providers on the cross-billing allegation without account-level corroboration.
- Confidence / evidence
- Limited · isolated or attribution-limited · reported incident coverage · community incident report
- Corroboration / limit
- The access interruption has secondary coverage; the cross-billing claim has only one unresolved direct forum report in this window. The billing attribution is unresolved, no broad quality sample is present, and neither receipt supports prevalence or causation.
Pro
$20
$20 API agent usage · Auto/Composer separate
Pro Plus
$60
$70 API agent usage · official docs
Ultra
$200
$400 API agent usage · official docs
Auto cap
Unpublished
official docs say generous included usage
Cost mechanics
Cursor is easiest to price when you separate the two pools.
Pool 2 · headroom
Auto/Composer Pool
Auto and Composer 2.5 draw from separate subscription headroom. Cursor publishes token classes and “generous” included usage, not the exact Auto/Composer cap.
Agent amplification
One prompt is not one billable operation.
In Agent mode, a narrow instruction can fan out into file reads, tool calls, commands, follow-ups with accumulated context, cache reads/writes, and multiple model calls. The risk is not a fixed multiplier.
High token usage threadCosts climbing reportAuto routing/usage report
Tier evidence
Pro / Pro Plus / Ultra, without inventing an Auto cap.
Pro
Cursor IDE plan
$20/mo
- API Budget
- $20 API agent usage
- Auto/Composer Pool
- Generous Auto + Composer usage; exact cap is not published.
- Overage control
- On-demand can be enabled after included usage; otherwise stop or upgrade.
Pro Plus
Cursor IDE plan
$60/mo
- API Budget
- $70 API agent usage
- Auto/Composer Pool
- Same published wording: generous Auto + Composer usage, no exact cap.
- Overage control
- Larger included API budget; on-demand and spend caps remain explicit controls.
Ultra
Cursor IDE plan
$200/mo
- API Budget
- $400 API agent usage
- Auto/Composer Pool
- Largest individual headroom before on-demand; same unpublished exact cap.
- Overage control
- Largest individual API budget before on-demand usage or a team/enterprise setup.
Choose Cursor when…
The IDE workflow is the value.
- IDE-first developers want agent loops, edits, context, cloud agents, Bugbot, and model selection inside one editor.
- Teams can manage risk with dashboards, spend alerts, seat types, and explicit limits.
- Buyers accept unpublished Auto/Composer exact headroom.
Avoid Cursor when…
Predictability beats IDE convenience.
- You need a published exact Auto cap.
- Finance needs strict chargeback or task-level attribution.
- You prefer predictable CLI/API metering with no routing ambiguity.
Teams migration evidence
Teams now has Standard/Premium seats and separate pools.
On June 1, 2026 Cursor announced separate Composer/Auto and Third-Party API pools, Standard at $40 monthly/$32 annual, Premium at $120 monthly/$96 annual, Premium 5× Standard usage for 3× price, and renewal migration starting July 1, 2026.
Source receipts
Flat dated receipts for every decisive claim.
| Date | Source | Claim used | Label | Link |
|---|---|---|---|---|
| 2026-06-24 | Cursor pricing | Individual pricing lists Pro at $20/mo and Teams Standard at $40/user/mo. | official pricing | open source |
| 2026-06-24 | Cursor usage and limits | Pro includes $20 API agent usage, Pro Plus $60/mo includes $70, and Ultra $200/mo includes $400; each also includes generous Auto and Composer usage. | official docs | open source |
| 2026-06-24 | Cursor Models & Pricing | Individual plans have two monthly pools: Auto + Composer for Auto/Composer 2.5, and API for selected-model or Premium routing at model API rates. | official docs | open source |
| 2026-06-24 | Cursor Models & Pricing | Auto pricing, Composer pricing, model API rates, Premium routing, Teams plans, Cursor Token Rate, and Max Mode pricing are documented as token/API-rate mechanics. | official docs | open source |
| 2026-06-24 | Cursor usage-based charges | On-demand usage must be explicitly enabled; it has separate invoices/line items, can use a spend limit, and additional usage is billed at API rates with no markup. | official docs | open source |
| 2026-06-01 | Cursor Teams pricing update | Teams moved to separate Composer/Auto and Third-Party API pools; Standard is $40 monthly/$32 annual, Premium is $120 monthly/$96 annual, Premium has 5× Standard usage for 3× price, and renewals migrate starting July 1, 2026. | official changelog | open source |
| 2026-01-23 | Cursor forum: Auto mode complaint | A user reported believing Auto mode was unlimited and then seeing usage exhausted, evidence of buyer confusion around Auto headroom. | user complaint/report | open source |
| 2026-04-22 | Cursor forum: Auto token-limit report | A forum response describes Auto as a shared Auto plus Composer budget separate from API, with visible token counts affected by router choices, caching, and per-token prices. | forum explanation/report | open source |
| 2026-04-09 | Cursor forum: costs climbing report | A user reported monthly Cursor cost rising from under $200 to over $400 without a clear workload change, plus prompts around $1 and stopped runs that still consumed spend. | user complaint/report | open source |
| 2026-04-07 | Cursor forum: high token usage | A forum response explains that Agent mode can turn tool calls and follow-up messages into separate API calls carrying chat context, system prompt, and tool results. | forum explanation/report | open source |
| 2026-05-30 | Cursor forum: Auto routing/usage report | A user reported similar work consuming quota faster after an update and cited a maintenance task using about 3.3M tokens in Auto agent mode. | user complaint/report | open source |