Cursor's June 2025 pricing switch left Pro users with surprise usage bills and mass refunds
Anysphere (Cursor) · Jun 16, 2025 · Coding agent
What happened
On June 16, 2025 Cursor replaced the Pro plan's 500-requests-per-month cap with a $20 monthly pool of frontier-model usage priced at API rates, with additional usage billed at cost. Users who had budgeted by request counts received large unexpected charges over the following weeks. On July 4 Anysphere published an apology titled 'Clarifying our pricing', admitted 'We missed the mark', and offered full refunds for any unexpected usage billed between June 16 and July 4.
Causal vector
Subscription pricing silently converted from request-capped to metered at-cost usage, with no hard spend ceiling on the pool
Source
Reported by Cursor blog - Clarifying our pricing (2025-07-04). Verified against the primary report.
A metered usage pool with no ceiling is an unbounded spend path by design. A hard monthly USD cap enforced outside the vendor's billing page turns a 'usage credit pool' back into the limit the old request cap promised - and it cannot be silently repriced in a blog post.
The six rule types that contain this class of failure
Per-transaction cap
Any single spend above your ceiling is BLOCKED before it moves.
Daily total
Cumulative spend across all agent calls, bounded per day.
Velocity limit
Stops runaway retry loops — the #1 cause of overnight losses.
Merchant allowlist
Only approved destinations can ever receive funds.
Category rules
Flag high-risk classes (crypto, infra, refunds) for review.
Approval threshold
Above a value, the action waits for a human.
Related incidents
Codex Windows app exhausted a weekly usage limit in about 62 hours across many parallel tasks
lty418905-blip (OpenAI Codex user) · Aug 5, 2026
Codex context auto-compaction loop rereads files, loses progress, and burns paid credits
edisonjoao1 (OpenAI Codex user) · Jul 24, 2026
OpenAI confirmed Codex background tasks were silently burning users' usage limits faster than normal
OpenAI (Codex) · Jun 30, 2026
Don't be the next entry
Every incident in this database is the result of trusting a prompt, a provider cap, or a human review cycle. sipi.bot replaces all three with one deterministic call. 90 documented failures, one control.