Cursor launches AI agent experience to compete with Claude Code and OpenAI Codex

Cursor Unveils Agent-First Interface to Transform Developer Workflows

Cursor announced the launch of Cursor 3, a redesigned AI coding platform that centers around AI agents capable of completing entire programming tasks autonomously1

1

. The product, developed under the code name Glass, represents the startup's response to the growing threat from Claude Code and OpenAI Codex, which have attracted millions of developers in recent months1

1

. According to Jonas Nelle, one of Cursor's heads of engineering, the profession has completely changed in recent months, making the company's previous product approach less relevant going forward1

1

.

The new AI agent experience features a chatbot-like interface where developers can type natural language descriptions of tasks they want completed, without writing a single line of code1

1

3

3

. Users can specify which large language models (LLMs) should carry out each task, and Cursor 3 generates the requested code along with demo videos showing how it works3

3

. Developers can view and manage all running AI coding agents through a newly added sidebar, creating what the company describes as a "unified workspace for building software with agents"2

2

.

Dual Agent System Balances Cloud Power with Local Control

What distinguishes Cursor 3 from standalone tools is its integration of cloud-based and desktop agents within the same development environment1

1

. Cloud-based AI agents have access to more hardware resources, allowing large numbers to run in parallel for time-consuming tasks3

3

. Desktop agents run slower but enable developers to locally open generated code, edit it manually, and run tests3

3

.

Cursor's cohead of engineering for Cursor 3, Alexi Robbins, demonstrated how users can prompt an agent in the cloud to build a feature, then review the generated code locally on their computer1

1

. This flexibility allows developers to switch between modes seamlessly—having cloud-based agents generate code before sending it to desktop-based agents for local editing3

3

. The platform also displays step-by-step overviews of each task completion, providing natural language explanations of every sub-step, highlighting errors, and offering screenshots3

3

.

Intensifying Competition Reshapes AI Coding Market

Cursor increasingly finds itself competing directly with leading AI labs for developers and enterprise customers1

1

. The company pioneered one of the first popular ways for developers to code using AI models from OpenAI, Anthropic, and Google, making it one of these companies' biggest AI customers1

1

. However, in the last 18 months, OpenAI and Anthropic launched their own AI-powered coding offerings through highly subsidized subscriptions that have pressured Cursor's business model1

1

.

Claude Code has reportedly captured up to 54% of the AI coding market, according to data from Menlo Ventures2

2

. OpenAI's recently released Codex 5.3 set new highs in benchmarking tests, with the company offering unlimited access to pull in more users2

2

. While Cursor remains popular among product and engineering teams, its position as the dominant player has eroded2

2

.

Proprietary Models and Pricing Changes Signal Strategic Shift

To reduce reliance on external providers, Cursor has invested in creating proprietary AI models like Composer 2, which is optimized for agentic tasks and balances speed, cost, and performance within the code editor4

4

3

3

. The model is more cost-efficient than several other LLMs supported by the AI coding platform3

3

.

However, the launch follows a bumpy rollout of Composer 2 last month, when users discovered the model was largely a licensed version of the open-source Kimi 2.5 model made by Moonshot AI2

2

. The lack of upfront disclosure has made some users wary of the company's tactics2

2

. Additionally, Cursor shifted from a flat subscription model to usage-based pricing in 2025, a change that has generated user dissatisfaction despite being positive for the company's sustainability4

4

.

Cursor, officially Anysphere Inc., has raised more than $3 billion in funding from Nvidia Corp., Google LLC, and other prominent backers3

3

. The startup is reportedly raising fresh capital at a $50 billion valuation—nearly double its valuation from a funding round last fall1

1

. As Nelle and Robbins noted, they simply want people using Cursor, regardless of which interface developers spend their time in1

1

. The battle to create the best AI coding tools may prove to be Cursor's most capital-intensive chapter yet1

1

.