
TL;DR
Cursor shipped a deliberate change on July 31 making the Usage page tokens-only for self-serve plans, removed the dollar Cost column, and zeroed per-request cost fields in the dashboard API - including for historical records. Staff confirmed the change is intentional and that the numbers are still tracked internally.
On July 31, Cursor changed its Usage page to show token counts instead of dollar amounts for self-serve plans, including Teams. The dollar Cost column disappeared from the table, the per-day spending graph went away, and the usage CSV stopped carrying cost data. The company's response in the thread made it clear this was not a bug: "This was an intentional change, not a temporary reporting issue. Usage reporting for self-serve plans, including Teams, is now token-based, so dollar values are no longer returned by the dashboard Usage endpoint."
The move landed the same way users track spend every day, and the pushback was immediate. If you budget an agentic coding subscription in dollars, tokens are a downgrade in every sense that matters. Here is what actually changed, what staff said, and what it means for anyone who watches their AI coding spend.
The change hit three surfaces at once:
get-filtered-usage-events endpoint now returns zeroed cost fields (chargedCents: 0, usageBasedCosts: "$0.00") for every event, including historical ones that previously returned real values.Cursor staff confirmed the endpoint behavior in the thread: "Because this is applied when records are read, historical results are affected, too." The period totals still exist internally, so the data is being tracked - it is just no longer exposed to the customer.
The one carve-out: enterprise plans still show dollar amounts. Staff explained the split as deliberate, because enterprise plans pool usage while individual and Teams plans mix included and on-demand usage. The stated rationale for the self-serve change: showing a dollar value on plan-covered usage confused people, since the number (computed at API rates) was often higher than what they were actually billed.
Dollar figures are not gone entirely. On-demand spending still appears on the Spending page, the CSV Cost column still carries dollars for on-demand rows, and Teams admins can see per-user on-demand totals under Members. What disappeared is the per-model, per-day, per-request dollar breakdown that self-serve users had for years.
From the archive
Aug 1, 2026 • 5 min read
Aug 1, 2026 • 6 min read
Aug 1, 2026 • 8 min read
Aug 1, 2026 • 7 min read
The community reaction split into two camps, and the skeptical one is loud.
The first theme: you budget in dollars, not tokens. Multiple users said the Usage page was effectively permanent open in their browser, refreshed after every model switch to see what a message cost. One described using it to compare models directly - sending a prompt through one model, watching roughly $0.32 land, switching to a frontier model and watching $2.56 land. That is a real engineering workflow: measuring model price/performance from your own workload, not from marketing pages. Tokens do not carry that signal, because cost per token varies by model, context, and caching.
The second theme: retroactive data removal is the line. Users who built reporting on the dashboard API - or exported CSVs daily for team budgeting - found the fields zeroed on historical records too. One user's summary captured the sentiment: "Transparency about what I'm being charged per request is not optional for a metered product - removing it retroactively breaks any independent cost tracking." Several noted that the aggregate billing totals do not replace per-model breakdowns, and that Teams admins now have no per-model dollar view at all on self-serve.
The third theme: trust. Thread participants described the change as user-hostile, and several said it pushed them toward alternatives that still show raw API-style costs. A few pushed back in the opposite direction, noting the old dollar figure included plan-covered usage at API rates and was never the actual bill - which is the exact confusion staff cited as the reason for the change. That defense went only so far: the common demand was a toggle or a clear split between included and on-demand usage, not the removal of the number.
Strip the product-decision debate away and two durable facts remain.
First, the models are metered, and the meter is now half-invisible. The API routes that let users reconcile what they actually spend on agentic work now return zeros. Any team that built a cost dashboard, a per-developer budget, or an automated alert on Cursor's usage API is now blind, and had no warning that historical data would be affected. That is a real operational risk for teams whose billing runs through a vendor's self-serve API instead of their own proxy.
Second, "Included" is not a number. Plan-covered usage is now a status word rather than a quantity, which makes it impossible to estimate how close you are to an on-demand bill without actively watching the Spending page. Anyone who used the Usage page to answer "is this model worth the marginal cost" has lost the cheapest instrumentation they had.
The timing matters too. Agentic coding tools are converging on the same accounting problem: usage-based models like Cursor's Ultra run on token pricing, subscription plans bundle generous included usage, and users keep asking one question - what is this actually costing me? That question does not go away because the answer is harder to see. If anything, opacity raises the value of the tools that answer it anyway: local TUIs that parse your own logs, spend-guardrail layers, and vendors that still return per-request cost fields on their APIs.
Cursor staff have asked users for feedback on the change, and the thread is the feedback. Whether the company keeps the token-only design or ships a toggle is still open. Either way, the takeaway for developers is the same as it has been all year: if your AI tool spend matters to you, keep an independent accounting of it, because the vendor's dashboard is not a contract.
Read next
GitHub Copilot switched to AI Credits billing on June 1 - here is what the change means for your team's budget, how Copilot Max fits in, and how costs compare to Claude Code and Codex.
7 min readCursor v3.11 introduces Side Chats for parallel agent conversations, Conversation Search across past sessions, and Cloud Agent Hooks for self-correcting loops. A practical guide to the new features released July 10, 2026.
6 min readComparing LLMs by token pricing alone can lead you to choose worse, more expensive models. Cost per task tells the real story.
6 min readTechnical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.
AI-native code editor forked from VS Code. Composer mode rewrites multiple files at once. Tab autocomplete predicts your...
View ToolThe original AI coding assistant. 77M+ developers. Inline completions in VS Code and JetBrains. Copilot Workspace genera...
View ToolCodeium's AI-native IDE. Cascade agent mode handles multi-file edits autonomously. Free tier with generous limits. Stron...
View ToolStackBlitz's in-browser AI app builder. Full-stack apps from a prompt - runs Node.js, installs packages, and deploys....
View ToolA concrete step-by-step guide to moving your development workflow from Cursor to Claude Code - settings, rules, keybindings, and the habits that transfer.
Getting StartedStep-by-step guide to building an MCP server in TypeScript - from project setup to tool definitions, resource handling, testing, and deployment.
AI AgentsContext-aware follow-up suggestions derived from git history.
Claude Code
OpenAI slashes GPT-5.6 Luna by 80% to $0.20/M input tokens, cuts Terra by 20%, adds Sol Fast mode at 2.5x speed, and rev...

A practical guide to routing between Claude Opus 5, Sonnet 5, Haiku 4.5, GPT-5.6 Sol/Terra/Luna, and Kimi K3 based on ta...

Tom Lockwood investigated the Bun Rust rewrite six weeks after it merged to main - 2,475 open PRs, no release tag, and c...

Ruff v0.16.0 ships 413 default rules (up from 59), Markdown code-block formatting, and a new ruff: ignore system. Here i...

A deep dive into why fully autonomous AI coding agents degrade codebases over time, and what context engineering can act...

Compare every verified Kimi K3 access route, including Moonshot, Together, Fireworks, Baseten, Modal, Vercel AI Gateway,...

New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.