Cursor has launched a new AI agent experience for its coding platform, putting the startup in direct competition with Anthropic's Claude Code and OpenAI's Codex for the first time, according to reporting by Wired.

Until now, Cursor occupied a relatively comfortable position: an editor built on top of models supplied by OpenAI and Anthropic. That arrangement is becoming harder to sustain as both model providers move aggressively into the agentic coding layer where Cursor has built its reputation.

Cursor now has to compete with OpenAI and Anthropic more directly than ever — the same companies whose models power its product.

The Agentic Coding Race Heats Up

The launch reflects a broader shift in how AI coding tools are evolving. First-generation assistants handled autocomplete and inline suggestions. The current wave — agents — can plan, execute, and iterate across entire codebases with minimal human intervention. Claude Code, released by Anthropic earlier this year, operates as a terminal-based agent that can read, write, and run code autonomously. OpenAI's Codex, relaunched as a cloud-based coding agent, works similarly. Cursor's new experience appears designed to match that ambition within its own editor environment.

For developers, the distinction matters. Editor-native agents like Cursor's can integrate directly with the workspace, file system, and project context a developer already has open. Terminal-based or cloud-based agents require more context-setting and often operate with less visibility into what the developer is actually doing.

What Cursor Is Building Against

Anthropic and OpenAI hold a structural advantage that makes Cursor's position delicate: they control the models. Cursor depends on API access to frontier models to deliver its experience. If those providers choose to prioritize their own agent products — through pricing, rate limits, or early access to new capabilities — Cursor's competitive position weakens without any direct action on its part.

The startup's counterargument has always been the quality of the developer experience. Cursor built a devoted user base by wrapping model capabilities in an interface that felt purpose-built for coding workflows, rather than grafted onto a general-purpose chat interface. The new agent launch appears to double down on that logic: that the environment in which an agent operates is as important as the model driving it.

Details on specific pricing tiers for the new agent features, or whether they fall within Cursor's existing $20-per-month Pro plan or require a separate add-on, were not confirmed in available reporting at time of publication.

The Startup Squeeze

Cursor's situation illustrates a structural tension that affects much of the AI application layer. Startups build valuable products on top of foundation models, grow a user base, and then face the risk that the model providers — flush with capital and increasingly product-focused — move into the same space. Cursor reportedly reached $500 million in annualized revenue earlier this year, a figure that demonstrates real commercial traction but also makes it a more visible target.

The response available to Cursor is to move faster on the product layer than the model providers can. OpenAI and Anthropic are, by nature, large organizations balancing research, safety, policy, and commercial product work simultaneously. A focused coding startup can, in theory, ship developer experience improvements faster. Whether that speed advantage holds as the model providers invest more in their own coding products is the central question Cursor's next chapter will answer.

For developers evaluating which agentic coding tool to adopt, the practical considerations are integration complexity, model quality, pricing, and workflow fit. Cursor's editor-native approach keeps it relevant for developers who want agent capabilities without leaving the environment where they already work. Claude Code and Codex appeal to developers comfortable with terminal or cloud-based workflows, or those already deeply integrated into Anthropic or OpenAI's broader tooling ecosystem.

What This Means

Cursor's new agent launch forces a direct comparison with the most well-resourced players in AI. Developers evaluating agentic coding tools now have a genuine three-way choice, and the deciding factor is likely to be workflow fit rather than raw model capability.