Coding agents
How Claude Code, Codex, Cursor and the others differ
In short
Claude Code (Anthropic) and Codex (OpenAI) both have sessions that keep running after you close the laptop (Claude Code, Codex). Claude Code’s Projects beta runs each task as its own thread in parallel and has let in its whole Pro and Max waitlist (post). Its desktop app gives each parallel task a nested session (post). In Codex, GPT-6.1 Sol has a default context of 500,000 tokens (post). Its Ultrafast mode costs $12 per million input tokens and $60 per million output tokens, while standard Sol stays at $2/$10 (post). Cursor now belongs to SpaceX. Its plans launched October 12 include Grok usage and bill every Claude request at API list prices (post). xAI’s $30/month Premium tier bundles SuperGrok, Cursor Pro and X Premium+ (post). At Universe’26, GitHub added Copilot Fusion, Automations, Projects, dreaming and a Security extension (post).
Latest changes: from December, Plus, Pro and Business usage in ChatGPT and Codex comes out of one shared allowance measured in tokens. Instant chat doesn’t count against it (post). OpenAI is doubling limits on the Pro 100, 200 and 500 tiers and Business Premium, and it cancelled the planned Pro 200 cut from 20x to 10x the Plus allowance (post). Google’s Android CLI plugin is now available for Claude Code and Codex, and Android Studio Rabbit 2 makes Bring Your Own Agent stable for over 40 ACP-compatible agents (post). From November 29, OpenAI may use Daybreak Blue sessions on personal accounts to improve its models even with the model-improvement setting off (post). Researchers say the embedded browsers in Codex and VS Code can still read any file on disk (post).
At a glance
| Agent | Company | What matters now | Price | As of | Source |
|---|---|---|---|---|---|
| Codex and ChatGPT shared usage pool | OpenAI | From December, Plus, Pro and Business usage in ChatGPT and Codex comes out of one allowance measured in tokens. Instant chat stays outside it. Balance and reset times show in Settings → Usage. | Plus, Pro and Business plans; credits can be bought when the pool runs out | 1 November 2026 | Read |
| Codex on Pro and Business Premium | OpenAI | Usage limits doubled on the Pro 100, 200 and 500 tiers and Business Premium. The planned Pro 200 cut from 20x to 10x Plus is cancelled. Extended limits for Dots stay "for now". | Pro 100, Pro 200, Pro 500, Business Premium | 29 October 2026 | Read |
| Codex (GPT-6.1 Sol) | OpenAI | Default context length of 500,000 tokens in Codex and ChatGPT Work, up from 272k. OpenAI plans to raise it further. | 12 October 2026 | Read | |
| GPT-6.1 Sol Ultrafast | OpenAI | Eight times faster than Sol Standard. Rolling out in ChatGPT Work, Codex and the API (service_tier: "ultrafast"). | $12 per million input tokens, $60 per million output tokens; Standard Sol $2/$10 | 8 October 2026 | Read |
| Claude Code Projects | Anthropic | Beta. One ongoing conversation in which Claude runs each task as its own thread in parallel. The whole Pro and Max waitlist is in; new sign-ups get in "as capacity allows". | Pro and Max plans | 9 October 2026 | Read |
| Claude Code desktop app | Anthropic | Each parallel task runs in its own nested session. Environments save a setup with an optional worktree per folder, and /btw opens side questions in a separate tab. | 14 October 2026 | Read | |
| Cursor | SpaceX | Plans launched October 12 include Grok usage and no Claude. Every Claude request is billed on top at API list prices, and the cheapest plan has no Claude at all. | $30 Premium plan with an included weekly Grok pool | 13 October 2026 | Read |
| The X Subscription (Cursor Pro bundle) | xAI | The Premium tier bundles SuperGrok, Cursor Pro and X Premium+, which share one pool of Grok usage. Third-party models in Cursor on Premium and above are paid for through top-ups at API list prices. | Premium $30/month; tiers from Plus at $8/month to Ultra at $200 | 13 October 2026 | Read |
| Cursor Computer Use | SpaceX | Reportedly operates other apps on your laptop, mobile simulators included, and runs on your own machine. Turned on on demand or with the "Computer Use" @-mention. | 23 October 2026 | Read | |
| GitHub Copilot | GitHub | Universe'26 features: Copilot Fusion, the Security extension, Copilot Automations, Copilot Projects and Copilot dreaming. | 29 October 2026 | Read | |
| Dots | OpenAI | The primary Dot is included in the plan and runs 24/7 without using the allowance. Codex tasks it opens draw usage as usual. | 1 October 2026 | Read | |
| Claude Code cloud sessions | Anthropic | Sessions keep running after you close the laptop. | 24 September 2026 | Read |
Show all (23)
| Agent | Company | What matters now | Price | As of | Source |
|---|---|---|---|---|---|
| Codex cloud | OpenAI | Keeps coding in the cloud after your laptop closes. | 29 September 2026 | Read | |
| Codex with open models | OpenAI | Opens up to Kimi K3 and GLM-5.3 Flash via Baseten. | 30 September 2026 | Read | |
| Android CLI plugin and Android Studio Rabbit 2 | Official plugin for Claude Code and Codex. Bring Your Own Agent is stable for over 40 ACP-compatible agents. | 28 October 2026 | Read | ||
| OpenCode | OpenCode | Open source coding agent. It says a Claude subscription now works inside it, while CellCog says Anthropic's SDK docs bar claude.ai login in third-party tools unless approved. | 30 October 2026 | Read | |
| DeepSeek Harness v0.2 | DeepSeek | Desktop app powered by DeepSeek models for work and coding tasks, with plugins and workspaces. Users report v0.2 runs on macOS and Windows but not on Linux. | 2 October 2026 | Read | |
| Muse Code | Meta | Out of beta, with an SDK and subscriptions. | 1 September 2026 | Read | |
| Air | JetBrains | JetBrains folds its agent tools into one brand, Air. | 22 September 2026 | Read | |
| Microsoft Copilot Code mode | Microsoft | Code mode in Microsoft Copilot, built on GitHub Copilot tech. | 25 September 2026 | Read | |
| Kiro | GPT-5.6 is available as an option. | 25 August 2026 | Read | ||
| Antigravity 2.0 | A Git panel and a terminal are built in. | 25 August 2026 | Read | ||
| Flint | Recursive | Open-weights coding model for agents: a mixture of experts with 39.8 billion parameters, about 2.5 billion active per token. 54.9% on Terminal-Bench 2.1 in Recursive's own evaluations. | Apache 2.0 open weights | 22 October 2026 | Read |
Milestones
- The 5-hour Codex cap returns for Plus accounts.
- Muse Code exits beta with an SDK and subscriptions.
- Anthropic keeps Claude in SpaceX-owned Cursor.
- Astra reaches Pro and Business plans before Plus.
- JetBrains folds its agent tools into one brand, Air.
- Claude Code sessions keep running after you close the laptop.
- Codex keeps coding in the cloud after your laptop closes.
- Codex opens up to Kimi K3 and GLM-5.3 Flash via Baseten.
- OpenAI launches Dots: the primary Dot is included in the plan and runs 24/7, but any Codex task it opens draws on usage.
- Ultrafast mode comes to GPT-6.1 Sol in ChatGPT Work, Codex and the API at $12/$60 per million tokens.
- The Claude Code Projects beta admits everyone on its Pro and Max waitlist.
- GPT-6.1 Sol’s default context in Codex and ChatGPT Work goes from 272k to 500,000 tokens.
- xAI’s The X Subscription bundles SuperGrok, Cursor Pro and X Premium+ into a $30/month Premium tier.
- Cursor’s new plans include Grok usage and bill Claude requests at API list prices.
- The Claude Code desktop app gives each parallel task its own nested session.
- Google opens Android Studio and Android CLI to Claude Code, Codex and other ACP-compatible agents.
- GitHub announces Copilot Fusion, Automations, Projects, dreaming and a Security extension at Universe’26.
- OpenAI doubles limits on the Pro 100, 200 and 500 tiers and Business Premium, and cancels the Pro 200 cut.
- OpenAI says ChatGPT and Codex will share one usage pool, measured in tokens, from December.
FAQ
Can I still use Claude in Cursor?
Yes, but it’s billed separately. Cursor’s plans launched October 12 include Grok usage and no Claude. On the $30 Premium plan, every Claude request is billed on top at API list prices, and the cheapest plan has no Claude at all. Current subscribers keep their included Claude usage until they switch (post). In xAI’s The X Subscription, third-party models stay in Cursor on Premium and above and are paid for through top-ups at API list prices (post).
Do ChatGPT and Codex share usage limits?
From December they will. Plus, Pro and Business usage in ChatGPT and Codex will come out of one shared allowance measured in tokens. Instant chat stays outside it, and more advanced models and complex tasks use more of it. The combined balance and reset times will show in Settings → Usage. When it runs out, you can wait for the reset, upgrade or buy credits (post).
Did OpenAI cut Codex usage on the Pro 200 plan?
No. OpenAI had announced a change for October 30 that would have dropped Pro 200 Work and Codex usage from 20x to 10x the Plus allowance. On October 29 it cancelled that cut and said it was doubling limits on the Pro 100, 200 and 500 tiers and Business Premium. Dot usage remains separate from default usage (post).
Can I use Claude Code or Codex with Android Studio?
Yes. Developers using Claude Code or Codex on Android projects can install Google’s official Android CLI plugin. Android Studio Rabbit 2 makes Bring Your Own Agent stable for over 40 ACP-compatible agents, which gives them the IDE’s project structure, Compose Preview, Emulator and Logcat. Fix with AI for memory leaks is still in the Canary build (post).
Does a Claude subscription work in OpenCode?
OpenCode says it does. The open source agent posted on X that a Claude subscription is now usable inside it, but the claim comes only from OpenCode’s own account. According to CellCog, Anthropic’s SDK docs bar claude.ai login in third-party tools unless Anthropic approves it (post).
Latest
- Cloud sessions in Claude Code now set up your repo
- LSP plugin made Claude Code 11% pricier on C++ fixes
- WIRED: Accenture worker’s Claude agent ran on DOD network
- Cloud-run agents in Antigravity start reaching Ultra and Pro
- Conductor joins Meta and shuts Conductor Cloud on Nov 20
- Claude, GPT, Gemini install Sentry and Playwright every time
- OpenAI to buy Pierre, whose diff tools already run in Codex
- Revolut moves 2,000 engineers to self-hosted GLM at $0.30/M
- OpenAI’s Pando cut monorepo push latency nearly 5x
- Claude API enforces thinking blocks and moves to Rust
