Teams with mixed coding workloads
Route quick UI edits, routine maintenance and difficult reasoning tasks without teaching every developer a changing model catalog.
Independent tool overview
Cursor Router at a glance
Cursor Router is the model-selection system behind Cursor's Auto mode. It classifies each agent request by task type and complexity, then chooses from an approved model pool according to a Cost, Balance or Intelligence objective instead of requiring the developer to select a model manually.
Visit the official Cursor Router site ↗
- Product type
- Per-request AI model router inside Cursor Auto
- Current plans
- Teams and Enterprise as of August 31, 2026
- Modes
- Cost, Balance and Intelligence
- Surfaces
- Editor, Agents Window, web, iOS, CLI and SDK
- Billing
- Routed model list price plus team token fee when applicable
Overview
What Cursor Router is
Cursor Router is designed around a simple observation: routine edits do not need the same expensive frontier model as a difficult, long-horizon debugging task. A classifier inspects each request and its context and routes easier work to a faster, less expensive model while reserving more capable models for harder work.
Users do not configure an individual route. They select Auto and an optimization mode: Cost prioritizes token spend, Balance is the default for new users and trades among quality, speed and cost, and Intelligence favors more capable models for complex tasks. Every mode charges the list price of the model that actually handles the request.
As of August 31, 2026, current Cursor documentation limits the new Router to Teams and Enterprise plans. It is on by default for Teams; Enterprise administrators must opt in. Cursor says individual plans will receive it later, but it has not published an exact date.
Use cases
Who Cursor Router is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Model-cost governance
Standardize an Auto mode and approved model pool while retaining the option to push harder tasks toward more capable models.
Fast-moving model markets
Let Cursor update the candidate pool as models change instead of pinning the organization to one provider for every request.
Capabilities
Core Cursor Router features
Per-request classification
A machine-learning classifier evaluates task type and complexity before choosing the model that handles an agent request.
Three optimization modes
Cost minimizes token spend, Balance targets a practical daily quality-cost tradeoff and Intelligence routes difficult work toward higher-capability models.
Managed candidate pool
Cursor currently documents routing across GPT-5.5, Claude Opus 5, Grok 4.5 and Claude Fable 5, with the pool subject to change as models evolve.
Admin model controls
Team administrators can enable Router by team or group, restrict modes, choose a default and allow or block underlying models.
Model visibility controls
The routed model is hidden by default, but administrators can expose its identity in request details for cost analysis and debugging.
Cross-surface availability
Auto routing works in Cursor's editor and Agents Window, the web and iOS apps, CLI and TypeScript or Python SDK.
SDK access
Developers can request model id auto-smart and set optimize_for to cost, balanced or intelligence after confirming availability for the team API key.
Process
How the Cursor Router workflow works
Step 1
Define the goal
Choose whether the team is primarily optimizing cost, everyday quality or difficult-task performance, and identify workflows where manual model selection remains necessary.
Step 2
Set the approved pool
Allow the required and recommended models. Grok 4.5 is required; blocking both GPT-5.5 and Claude Opus 5 also disables the current Router.
Step 3
Select Auto and a mode
Choose Cost, Balance or Intelligence in the model picker, or set the equivalent optimize_for parameter through the SDK.
Step 4
Inspect quality and spend
Use the Spending dashboard and request-level model visibility to compare routed-model cost, completion quality, retries and time-to-accepted-change.
Step 5
Keep escape hatches
Let developers pin a specific model for reproducibility, regulated work or tasks where the router repeatedly makes a weak choice.
Cost
Cursor Router pricing and free plan
Cursor Router has no separate add-on price, but currently requires Teams or Enterprise. Auto requests bill at the routed model's per-token list price and consume the relevant Cursor Models or Other Models usage pool. On Teams and Enterprise, third-party models add a $0.25 Cursor Token Rate per million input, output and cached tokens. Included usage does not roll over; on-demand charges apply only when enabled.
Teams Standard
$40/user/month
Self-serve team plan with Router enabled by default and a standard included-usage allowance.
- Auto bills at the routed model's list price
- $0.25 per million-token Cursor Token Rate on third-party models
- First-party Grok and Composer models are exempt from the token fee
- On-demand usage can be enabled
Teams Premium
$120/user/month
Team plan with five times the Standard Agent limits and the same Router controls.
- Per-user usage pools
- Higher included allowance
- Third-party model and Cursor Token Rate rules still apply
Enterprise
Custom
Enterprise plan with pooled usage, advanced controls and administrator opt-in for Router.
- Custom commercial terms
- Model allow and block controls
- Invoice, SCIM and advanced governance options
Legacy Enterprise Auto Cost
Temporary through September 7, 2026
A short-lived flat token schedule for existing Enterprise Auto Cost traffic, regardless of the routed model.
- $1.25 per million input and cache-write tokens
- $0.25 per million cache-read tokens
- $6.00 per million output tokens
- Scheduled to end after September 7, 2026; verify current terms before relying on it
Pricing checked . Check current pricing at the source ↗
Assessment
Cursor Router strengths and limitations
Where it stands out
- Can reduce needless frontier-model spend on routine coding work without asking users to memorize provider pricing and strengths.
- Optimization modes give teams a simple policy lever instead of a large matrix of per-model rules.
- Administrator controls support organizational defaults, model restrictions and optional routed-model transparency.
- The same routing policy can follow work across interactive Cursor surfaces and programmatic SDK calls.
- Request-level billing exposes the actual model cost rather than charging one opaque flat rate for every routed request.
What to consider
- The new Router is currently limited to Teams and Enterprise; individual-plan rollout is promised but not dated.
- Users cannot choose the exact model for an individual Auto request, and the candidate pool can change, which reduces reproducibility compared with a pinned model.
- The routed model is hidden by default, making quality or cost debugging harder unless an administrator enables visibility.
- Balance and Intelligence consume usage faster than Cost, and third-party routes add Cursor's $0.25-per-million-token fee for Teams and Enterprise.
- Blocking Grok 4.5 disables the current Router, while blocking other recommended frontier models can degrade or eliminate routing quality.
- Cursor's claim that Intelligence delivers roughly 20% to 30% higher quality is a vendor claim and will not translate uniformly across repositories or tasks.
- Automatic routing can optimize model selection but cannot fix ambiguous prompts, broken environments, weak tests or unsafe generated changes.
Compare
Cursor Router alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Miscellaneous
Router
Ramp Router is a model-routing alternative for general AI API traffic when routing must sit outside Cursor and serve multiple applications.
Explore Router →Coding
Cursor Agents
Cursor Agents with a manually selected model are the direct alternative when reproducibility or explicit provider control matters more than automatic routing.
Explore Cursor Agents →Coding
Codex
Codex is an alternative coding-agent ecosystem for teams that prefer OpenAI's managed model and agent stack rather than Cursor's cross-provider routing.
Explore Codex →Questions
Cursor Router FAQs
What is Cursor Router?
It is the classifier and routing system behind Cursor Auto. It evaluates each coding-agent request and selects a model according to a Cost, Balance or Intelligence objective.
Which plans include Cursor Router?
As of August 31, 2026, Cursor documents it as available on Teams and Enterprise. Individual plans are expected later, but no exact rollout date is published.
How much does Cursor Router cost?
There is no separate Router fee. Each request is billed at the selected model's token rate and draws from the corresponding usage pool. Teams and Enterprise add a $0.25 per million-token fee when Router selects a third-party model.
What is the difference between Cost, Balance and Intelligence?
Cost prioritizes lower token spend, Balance targets a daily-driver tradeoff among quality, speed and cost, and Intelligence favors more capable models for difficult work and typically consumes usage faster.
Can I see which model Cursor Router selected?
Not by default. A team administrator can change model visibility to Displayed so users can inspect the routed model in request details.
Can administrators block models?
Yes, but the current Router requires Grok 4.5, and blocking too many frontier candidates can reduce quality or disable routing. Test policies before broad enforcement.
Can I use Cursor Router through an API?
Yes. Cursor's TypeScript and Python SDKs expose model id auto-smart with optimize_for set to cost, balanced or intelligence, provided the team's API key has access.
Bottom line
Our Cursor Router verdict
Cursor Router is a sensible cost-and-complexity abstraction for teams already using a varied model mix. Balance is the most reasonable pilot default, but the real proof should be accepted-change rate, retry cost and developer time on the team's own repositories. Keep model identity visible during evaluation and preserve manual selection for reproducible or high-risk work.
Visit Cursor Router website ↗Similar Coding Tools
View all Coding AI tools →
Laguna S 2.1
Laguna S 2.1 - Poolside's open-weight coding model with a 1M-token context
MAI-Cyber-1-Flash
MAI-Cyber-1-Flash - Microsoft's new cybersecurity model for securing codebases
LM Studio Bionic
LM Bionic - LM Studio's local-first agent that codes and edits docs using open models
Muse Code
Muse Code - Meta's terminal-based AI coding agent
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.