AI Coding Agents: Claude Code, Cursor, Copilot, Codex — 2026-09-12
OpenAI has opened a public test for its Codex Agents API, promising no extra fees beyond model usage, while GitHub Copilot faces continued developer backlash over its recent shift to token-based billing. Meanwhile, Japanese media reports on the launch of an official Unity plugin for Claude Code, signaling deeper integration into game development workflows.
AI Coding Agents: Claude Code, Cursor, Copilot, Codex — 2026-09-12
Top developments
OpenAI Launches Public Test for Codex Agents API
On September 10, 2026, OpenAI released the underlying runtime framework for Codex as a public test API, allowing developers to apply for access. The official stance emphasizes that there are no additional fees for the framework itself; users only pay for the actual model tokens and tools utilized. This move aims to lower the barrier for building custom agentic workflows while maintaining the existing consumption-based billing model.
GitHub Copilot’s Billing Shift Accelerates Competitor Adoption
GitHub Copilot’s mid-June switch from flat-fee subscriptions to metered, token-based credits has sparked significant backlash among its 1 million+ paying users. Recent reports indicate that this friction is driving teams to evaluate open-source alternatives and rivals with fixed pricing structures. The shift highlights the growing tension between sustainable AI infrastructure costs and developer expectations for predictable software expenses.
Official Unity Plugin Released for Claude Code
Japanese tech outlet Mado no Mori (Window Forest) reported on September 10 that an official Claude Code plugin for Unity has been released. The plugin integrates 29 specific skills designed to assist game developers directly within the Unity engine environment. This release marks a significant expansion of Claude Code’s utility beyond general software engineering into specialized verticals like game development.
Local view
Chinese Media Focus on Cost Structures Chinese tech community platform Smzdm (What's Worth Buying) published an analysis on September 12 discussing the nuances of OpenAI’s "no extra fee" claim for the Codex API. The article clarifies that while the API interface itself is free, the cost model remains strictly tied to token consumption, warning developers not to misinterpret this as a free service.
Japanese Press Highlights Industry Integration In addition to the Unity plugin news, Japanese outlet SHIFT AI TIMES updated its guide to Claude Code pricing and limits on September 11, reflecting the tool's growing adoption in the local enterprise sector. The article breaks down the $20 Pro and $100-$200 Max tiers, emphasizing the importance of understanding rate limits for terminal-based coding tasks.
Context & numbers
Adoption and Trust Metrics Recent data compiled by Uvik Software indicates that while 84% of developers now use AI coding tools daily, only 29% express full trust in the output. GitHub Copilot remains the leader in total user base with approximately 20 million users, though it faces stiff competition from Cursor, which has crossed $1 billion in annualized revenue.
Pricing Landscape The market has largely shifted away from flat-rate subscriptions. Current pricing models for major agents include:
- GitHub Copilot: Pro at $10/mo ($15 credits), Max at $100/mo ($200 credits).
- Claude Code: Pro at $20/mo, Max at $100-$200/mo.
- Cursor: Free tier available; paid plans range up to $200/mo for high-volume users.
On the radar
- Limit Change Ledger: Digital Applied continues to track the "38 dated changes" to paid AI coding plan limits across nine vendors in 2026, noting frequent discrepancies between vendor marketing and effective rate limits.
- ACP Registry Expansion: The Agent Client Protocol (ACP) registry, launched by Zed in early 2026, is seeing increased adoption with major tools like Claude Code, Codex CLI, and GitHub Copilot CLI registered, potentially standardizing how IDEs connect to these agents.
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
