What's New in Cursor in 2026

The table starts with the August releases, then covers the June and July releases around Compile. Each row gives the date and what changed. For setup, pricing, privacy or model choice, follow each topic's own guide, because a launch note is not permanent documentation.

Cursor's release notes and announcements say what changed. They don't say how a team should act on it. The steps below are our suggested order, and the next section lists what to watch out for along the way.

1. 1Start with the job: code hosting, always-on cloud agents, persistent cross-app work, model choice, cloud setup, mobile review, integrations, plan access or admin policy.
2. 2Use the dated entry to learn what changed, then open that topic's own guide for current behavior and limits.
3. 3Read the primary release page and current documentation before you quote how a feature works.
4. 4Update prompts, policies, training and pricing examples together, so your team learns one current version.
5. 5Recheck dates and price claims before you use them in sales, security or procurement conversations.

Each item below is either a fact about the release, such as its beta status or plan access, or our advice for using it safely. Plan around each one, and recheck the facts when Cursor ships an update.

• Keep these names apart: Grok BotAn early AI teammate product from SpaceXAI and Cursor that works across apps on a persistent cloud computer and asks for approval on selected actions; launch materials describe it as beta, but Cursor says Beta Services terms do not apply. Press Enter for the full definition. is a teammate product, Grok 4.6A Cursor and SpaceXAI model for complex coding and knowledge work, still in the Cursor model pool and succeeded by Grok 4.7; it is a selectable model and is separate from the Grok Bot product. Press Enter for the full definition. is a model, BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. is pull-request review and Origin is code hosting. Each has its own controls.
• Subscriptions are cloud-agent only for now, per the August 19 changelog. They are not a local agent feature.
• Grok BotAn early AI teammate product from SpaceXAI and Cursor that works across apps on a persistent cloud computer and asks for approval on selected actions; launch materials describe it as beta, but Cursor says Beta Services terms do not apply. Press Enter for the full definition. requires cloud data storage and does not support Privacy Mode (Legacy)A Cursor data setting that blocks Grok Bot because the product requires cloud data storage. Press Enter for the full definition.. Its Bots share one user computer's files and sign-ins.
• Grok 4.6A Cursor and SpaceXAI model for complex coding and knowledge work, still in the Cursor model pool and succeeded by Grok 4.7; it is a selectable model and is separate from the Grok Bot product. Press Enter for the full definition. launch pricing is time-bound where Cursor labels it a one-week promotion.
• Origin is still an early beta, not GA. Cursor has published no general-availability date, standalone pricing or self-hosting.
• Cursor for iOS has an official product page (cursor.com/mobile) and needs a plan that includes cloud agents. Treat Android as planned and web-only until Cursor ships it.
• The larger model is training work until Cursor announces that it is available.
• Keep older BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. and pricing figures apart from the June 2026 release figures.
• Never quote a Cursor Router cost figure without naming its baseline. Cursor's Router docs say Cost mode bills as a Cursor-model request, while Balance and Intelligence bill at the list price of the model each request is routed to. The help center's Router page says all Auto modes bill at the routed model's list price, so name the source you rely on.
• Check release mechanics against Cursor's changelog and blog pages.
• Customer stories show context. They don't replace product documentation.

Each topic in this release list has its own page, with more detail and its own sources. Use the table to go straight to the one you need.

There are two ways, and a Cursor presenter recommended both once a chat passes roughly 80% of its context window. In a new chat, @Chats brings in context from a prior conversation, and the presenter said it condenses that context and drops completed external and MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. tool calls, so the new thread starts from the gist. The second way is to fork the chat, which copies a prior chat's context into a new chat.

Because more AI-generated code means more code to review, and a presenter in Cursor's data science workshop said that makes review the next bottleneck. In the demo, a review button at the bottom of the chat listed every edit the agent had made, including the edits that built a chart.

Presenters in three Cursor workshops flagged these directions as forward-looking, not released. Treat them as signals, not as features you can use today, and check the changelog for any that have shipped since.

Start with the changes that act on you whether or not you act on them. Cursor Router shipped on July 22 and is on by default for Teams plans. It routes Auto per request. Cursor's Router docs say only Cost mode keeps classic Auto pricing, and Balance and Intelligence bill the model that served each request, plus the Cursor Token Rate on eligible third-party models.

Nobody on the team has to change a setting for the billing basis to change underneath them.

On Enterprise, the same feature works the other way. The router stays off until an admin turns it on, so Enterprise has a decision waiting while Teams has a default to check. Enterprise model access controls interact with it too. The router routes around blocked models, and Cursor says blocking too many reduces routing quality and can disable routing altogether.

Sort them by what each release asks of you. Grok BotAn early AI teammate product from SpaceXAI and Cursor that works across apps on a persistent cloud computer and asks for approval on selected actions; launch materials describe it as beta, but Cursor says Beta Services terms do not apply. Press Enter for the full definition. and Origin launched as early betas with plan gates, and both need cloud data storage, so legacy privacy mode rules them out. Decide your data and access boundary before you try either one. Google Workspace pluginsSeparate Cursor plugins for Gmail, Google Drive and Google Calendar that let an authorized agent read or change workspace data. Press Enter for the full definition. can act in Gmail, Drive and Calendar, so permission design is part of their setup.

Gemini 3.7 FlashGoogle's previous speed-tier model in Cursor, hidden by default and succeeded by Gemini 3.8 Flash at the same published rates. Press Enter for the full definition., Cloud Agent BuildsPrepared Cloud Agent environments built in the background, with the last successful build kept active when a newer build fails. Press Enter for the full definition., iPad review, Subscriptions, Custom modes and /goal are opt-in capabilities to evaluate against a concrete job. Cursor StartAn India-only legacy individual plan at ₹649 a month centered on Cursor's own models, with cloud and mobile access but without Other Models, Auto, Bugbot, Automations or the Agent SDK; Cursor names Plus as the comparable plan on The X Subscription. Press Enter for the full definition. sat in that group when it launched in July, and it is now a legacy plan. The larger from-scratch model remains an announcement, so don't plan around it yet.

That gives four buckets: default-on, gated beta, opt-in and announced. The opt-in bucket gets most of the reading time, while billing, privacy and approval surprises usually hide in the default and gated buckets. The August 19 harness changes are mostly opt-in, with one default worth knowing: cloud agents now subscribe to the pull requests they create and keep driving them, fixing CI and answering bot comments without being asked again.

Because Cursor has published comparisons against three different baselines, and the router is a saving or a spend depending on which one describes your team. Against Cost mode, Cursor's Router docs give no multiple and say only that Balance and Intelligence use your usage limits faster.

Against routing every request to Opus 4.8, Cursor reported early-access customers paying 30 to 50% less on Auto-routed requests, with Balance about 36% lower. The widely repeated 60% figure belongs to a third baseline: it is measured against Fable 5, not Opus 4.8. Cursor's August 6 post on how the router works updated both comparisons, to 41% lower cost than Opus 4.8 for Balance and 68% lower cost at Fable-level satisfaction for Intelligence.

So a team that has lived on Auto's bundled pricing should read the router as an increase, and a team that pinned an expensive model for everything should read it as a decrease. The direction depends on what the last invoice was built on.

Intelligence is the easiest mode to misread. Cursor's own numbers put it at about 15% higher satisfaction than Opus 4.8 at close to the same cost: $6.76 against $7.34 per commit. So on pinned Opus 4.8 traffic, the saving comes from Balance.

Budget a 60% cut on that migration and you will be wrong by a wide margin.

Separate the claims that expire from the ones that don't. A mechanic, such as what /in-cloud does or what a side chat inherits, holds until Cursor changes the feature. A number, such as a review time, a cost multiple or a seat price, expires quietly. A status word like beta or waitlist expires the moment someone ships. This tracker carries its own checked date instead of the site-wide one, because being current is its main claim.

Status words deserve the most suspicion, in both directions. Origin went from waitlist to early beta on August 17, with no standalone pricing or self-hosting, and with its agent-native features marked as coming soon. A slide that treats it as a finished GitHub replacement makes a promise nobody can keep, and one that still calls it a waitlist is out of date.

Platform lists drift too. Grok BotAn early AI teammate product from SpaceXAI and Cursor that works across apps on a persistent cloud computer and asks for approval on selected actions; launch materials describe it as beta, but Cursor says Beta Services terms do not apply. Press Enter for the full definition. launched in August with macOS, Windows and iPhone clients, and Cursor's docs now also list Linux and Android apps. Full pull-request review on iPad has held since July, on plans that include Cloud AgentsAgents that run in a Cursor-managed virtual machine, check out the repo, do the work and open a pull request, then shut down, with no load on your laptop. Press Enter for the full definition..

Small releases are where details bite, and side chats are a good example. They arrived in 3.11 and go one level deep, because a side chat can't spawn another. They are local-only for now, with Cloud AgentsAgents that run in a Cursor-managed virtual machine, check out the repo, do the work and open a pull request, then shut down, with no load on your laptop. Press Enter for the full definition. support marked as coming soon. Any one of those limits can break a workflow built on the feature.

Keep the habit small. Read the release note for the surface you own instead of a summary of all of them, then update the prompt, the policy and the pricing example together. A training deck that teaches the new mechanic on last quarter's pricing confuses a new hire more than one that is out of date all the way through.