Cursor Composer (Composer 2.5): an ICT teacher's prep tool, not a school platform provider
Composer 2.5 launched on 18 May 2026 as Cursor's own agentic coding model at $0.50 / $2.50 per 1M tokens for the standard variant, with an equally capable Fast variant at $3 / $15 that is the in-product default. It runs only inside Cursor's own products, exposes no inference API that can be called as a model endpoint, and was not designed for classrooms or students. This page explains how an ICT teacher can use it and why it does not appear in Edor.ai's provider list.
Not applicable to a school platform. Composer is a coding-specific model built into the Cursor editor, and the documentation states it is available only inside Cursor's own products, with no inference API that can be called as a model endpoint. Cursor does offer a Cloud Agents API and SDK, but those drive coding agents inside Cursor's environment rather than exposing a chat model a learning platform can call, and it is not designed for classrooms or students. Teachers can use Cursor as a tool when teaching programming, but the platform's model choice belongs among the other providers listed here.
Specifications
- Vendor
- Cursor (Anysphere) · United States
- Representative model
cursor-composer-2-5- Released
- 2026-05-18
- Context window
- Not published
- Max output
- Not published
- Modalities
- Text
- Open weights
- No
- API pricing
- $0.5 / $2.5 — USD per 1M tokens (input / output)
- Free tier available
- Yes
Composer 2.5 is a practical lesson-prep tool for ICT teachers; the mistake in procurement is assuming "the school bought Cursor" equals "the school AI platform has a programming model" — they are entirely different product categories, and the latter cannot be wired into Edor.ai.
Cursor is a code editor built by Anysphere, and Composer is the agentic coding model Cursor trains itself. The 2.5 release on 18 May 2026 pushed token pricing down to $0.50 / $2.50. This page does not answer "which provider should the platform use". It answers "should a teacher who teaches programming install this tool" — yes for preparation, no as a student chat or school platform model source. That boundary confuses management committees because both sit under the word "AI"; the platform question is about governed pupil access, while Cursor is an individual developer environment with no student audit trail.
What Cursor is actually building
Composer was never a general assistant. It was trained to edit files in a real project, run terminal commands, and keep iterating until tests pass. Unlike GPT or Claude, its strength is a visible long-running coding process — step-by-step edits, errors, fixes — which projects well as a lesson in debugging thinking.
The documentation states plainly that Composer 2.5 is available only inside Cursor — desktop, web, CLI and mobile. The Cloud Agents API and Cursor SDK can specify Composer to drive coding agents, but that edits repos and opens pull requests. It is not an endpoint where you send a pupil prompt and get a reply. A school platform needs the latter, and Cursor does not offer it.
Models a school procurement will encounter
| Model | Role | Context | Price per 1M tokens (input / output) |
|---|---|---|---|
| Composer 2.5 standard (18 May 2026) | Cursor's own agentic coding model; Cursor-only | — | $0.50 / $2.50 |
| Composer 2.5 Fast | Equally capable but faster; in-product default | — | $3 / $15 |
| Composer 2 | Predecessor to 2.5; RL-trained on long-horizon coding | — | — |
| Cursor Grok 4.6 / 4.5 (12 Aug 2026) | Shares Cursor Models usage pool with Composer | — | — |
Individual plans (September 2026): Hobby free, Pro $20/month, Pro+ $60/month, Ultra $200/month. Teams Standard $40/user/month, Premium $120/user/month; Enterprise custom-quoted.
Where the money actually goes
Individual subscription: A teacher preparing lessons alone — Pro at $20 a month is usually enough. Usage sits on the Cursor account, outside the Edor.ai annual fee.
Teams / Enterprise: Department sharing is seat-based, plus the Cursor Token Rate — third-party model calls (e.g. GPT) carry an extra $0.25 per 1M tokens on top of the model's own price; Cursor first-party models including Composer are exempt.
The Fast default trap: Standard $0.50 / $2.50 versus Fast $3 / $15 is a sixfold gap, and the default route is Fast. An ICT panel head told a colleague: "I thought we were on the cheap tier — the bill showed a whole evening of trial examples on Fast."
Composer's cheap tokens buy coding work, not the full teaching stack — composition feedback, knowledge-base retrieval and Cantonese speech are all outside its scope.
How it fits inside Edor.ai
Cursor Composer is marked not applicable on the platform — not "not supported yet", but no chat-model interface to connect. Edor.ai natively supports OpenAI, Azure OpenAI, Anthropic, Poe and local Ollama, and a learning platform's model choice should be made among those five.
If an ICT teacher wants Cursor for preparation or classroom demonstration, that is a separate teacher-tool decision: accounts, payment, privacy mode and student age terms all sit with the school. We have written the practical steps in the practical Cursor guide for educators. None of that replaces the five native providers Edor.ai actually wires in for pupil-facing work.
Classroom fit and hard limits
Teachers: preparation and demonstration for ICT or computer literacy — working teaching examples, lesson-by-lesson project splits, debugging exercises with deliberate errors left in. To demo a class library borrowing program, describe the requirement in plain language, get a working version, then break it into three lessons' worth of steps.
Students: not directly. Pupils who should experience AI-assisted programming belong in a gated, monitored environment. For the underlying ideas, start with the LLM Classroom and prompt engineering.
IT coordinators: treat it as individual teacher productivity, not part of the school system; watch account ownership, payment method, and keep sensitive school data out of unassessed development tools.
Does not fit: wiring it into Edor.ai as a chat model; letting pupils install Cursor for homework; using it for Chinese composition feedback; assuming the Cloud Agents API equals an OpenAI-compatible endpoint.
Paste this into Cursor for a three-lesson Python intro — then run the program yourself before projecting it:
I am a Hong Kong secondary ICT teacher preparing a three-lesson introductory Python project for Secondary 3: a small program that records class library borrowing and returns.
Output: a complete working program using only the Python standard library, a teaching sequence across three lessons, one five-minute exercise per lesson, and a list of common mistakes with hints.
Constraints: variables and comments in English, teaching notes in Hong Kong English conventions; use no syntax beyond Secondary 3 level; if an approach differs between Python versions, mark it "teacher to verify".
The trade-off for schools
- If you want programming support and general teacher tools inside the school platform, choose among native providers such as OpenAI or Anthropic — do not wait for Composer to open a chat API.
- If you approve Cursor for ICT lesson prep, set separate account and payment rules and train teachers to switch to standard rather than the default Fast tier, avoiding a sixfold bill surprise.
- If pupils should experience AI-assisted programming, do it inside a gated platform where the teacher sets how far it opens up, rather than handing students a professional developer tool directly.
Several lines from the same vendor
Most vendors keep a flagship, a workhorse, a lightweight and a reasoning line running at once, and the prices can differ tenfold. Picking the wrong line is a more common mistake than picking the wrong vendor.
| Model | Tier | Released | Context window | In / Out | Notes |
|---|---|---|---|---|---|
| Composer 2.5 | Coding | 2026-05-18 | — | $0.5 / $2.5 | Cursor's own agentic coding model at $0.50 / $2.50 per 1M tokens standard; an equally capable Fast variant costs $3 / $15 and is the in-product default. |
| Composer 2 | Coding | 2026-02 | — | — | The predecessor of 2.5 and the first Composer trained with reinforcement learning on long-horizon coding tasks. |
| Cursor Grok 4.6 / Grok 4.5 | Coding | 2026-08-12 | — | — | xAI models that share the Cursor Models usage pool with Composer; on team plans only first-party models are exempt from the Cursor Token Rate. |
Questions that keep coming up
What is cheap is the token price for coding work. A school platform needs an inference API you can call as a model endpoint — send a pupil prompt, get a reply. Composer runs only inside Cursor products. The Cloud Agents API and SDK drive coding agents that edit repos, not a chat endpoint, so there is nothing to wire into Edor.ai.
As of September 2026 the individual plans are Hobby free, Pro at $20 a month, Pro+ at $60 a month and Ultra at $200 a month. For a teacher planning and trialling lessons, Pro is usually enough. For a whole department, look at Teams — Standard at $40 per user per month and Premium at $120 per user per month.
Standard is $0.50 / $2.50, Fast is $3 / $15 — a sixfold gap — and Fast is the product default. A teacher who never switches assumes the cheap tier and gets a bigger bill than expected. Department procurement should put that in training.
On Teams and Enterprise plans, third-party model calls carry a Cursor Token Rate of $0.25 per 1M tokens on top of that model's own price. Cursor first-party models including Composer are exempt. If a department mixes Composer and OpenAI, price both lines separately.
We would not advise it. Cursor is a professional developer tool with no teacher monitoring, no hint-level ceiling and no audit trail of student conversations, and it was not designed for minors. Students meeting AI-assisted programming should do so inside a gated platform where the teacher sets how far it opens up.
Prices and specifications here were checked on 2026-09
- Cursor — Composer 2.5 launch post (date and pricing)
- Cursor Docs — Composer 2.5 model page
- Cursor Docs — models and pricing (usage pools and the Cursor Token Rate)
- Cursor — plan pricing page
- Cursor — Composer product page
- Cursor Docs — Cloud Agents API (the programmatic interface that can specify Composer)
- · Prices, features and specifications follow the official documentation linked above. Vendors change them without notice, so verify once more before you buy.
- · Product names and trademarks belong to their respective owners. Edor.ai has no partnership, agency or sponsorship arrangement with these companies.
- · This is an independent write-up for educational purposes. It is not procurement advice and not legal advice.
- · For anything involving student personal data, assess it against your school policy and the Personal Data (Privacy) Ordinance (PDPO) before you roll it out.
Related reading
- Cursor for Teachers (September 2026): Pricing, the 18-Year Line, and a Debugging Lesson You Can Actually Run
Cursor is Anysphere's AI code editor. As of September 2026 the plans are Hobby free, Pro US$20/month, Pro+ US$60, Ultra US$200, Teams US$40/user and Teams Premium US$120/user. This guide covers the two usage pools and what Composer 2.5 really costs, the Cursor Token Rate surcharge on Teams, what the 18-and-over terms mean for student accounts, and a full walkthrough of using it to teach debugging on the projector.
- Comparing 15 LLM Providers in 2026: Prices, Context Windows, and the One Decision a School Actually Makes
One set of school usage assumptions applied to September 2026 official pricing from OpenAI, Anthropic, Google, xAI, Meta, Mistral and Cursor Composer, plus DeepSeek, Qwen, Kimi, GLM, MiniMax, Doubao, Hunyuan and ERNIE. Capability gaps are narrowing while price gaps widen, which means the real decision is not which model to use but who absorbs the volatility and what a generational change costs.