Choose Codex for OpenAI's agent in the terminal, ChatGPT or Codex Cloud, included with ChatGPT plans, and Cursor for an AI editor that switches between OpenAI, Anthropic, Google and other models. Codex's IDE extension also runs inside Cursor, and both read AGENTS.md.[1][2][3][4]
Codex opened 31 of the 866 agent-written pull requests we merged from August 12 to October 4, 2026, and they merged a median 37 minutes after opening; Cursor opened none.
Measure both with Zest:
Codex and Cursor sessions side by side, per engineer
The model, skills and MCP tools behind each session
Claude CodeFable 5.1 (max) (with fallback)70.4 · Coding Agent Index
CodexGPT-6 Astra (max)67.0 · Coding Agent Index
Muse CodeMuse Spark 1.3 (xhigh)64.2 · Coding Agent Index
Grok BuildGrok 4.5 (high)64.1 · Coding Agent Index
Kimi Code CLIKimi K362.6 · Coding Agent Index
OpencodeGemini 3.8 Flash (high)61.2 · Coding Agent Index
Cursor CLIComposer 2.5 Fast38.3 · Coding Agent Index
Demo setups use current model families. The pairings above are the combinations benchmarked by Artificial Analysis; rankings measure performance, not popularity.
Illustrative daily trends · Other totals: last 30 days. Build Hours: this week, model-estimated activity including overlap; not engineer hours.
USED BY ENGINEERING TEAMS AT
AT A GLANCE
Codex vs Cursor, side by side
What each tool documents, row by row.
Codex vs Cursor, side by side
Codex
Cursor
Team seats for 5 engineers
$100/mo — ChatGPT Business at $20 a seat, billed annually ($25 monthly).[2]
$160/mo — Cursor Teams Standard at $32 a seat, billed annually ($40 monthly).[5]
Where it runs
Codex CLI, an IDE extension (VS Code, Cursor, Windsurf), the ChatGPT desktop app, Codex Cloud and @codex review on GitHub.[1]
The Cursor editor (Agent pane and Agents Window), the Cursor CLI, Cloud Agents in VMs, and JetBrains through ACP.[6]
Models
OpenAI GPT models; local clients can point at a custom Responses-API provider such as Ollama.[7][8]
OpenAI, Anthropic, Google, SpaceXAI (Grok) and more, plus Cursor's own Composer; Auto routing and bring-your-own keys.[3]
Project instructions
AGENTS.md (global ~/.codex, then repo root down to the working directory) and AGENTS.override.md.[9]
.cursor/rules, team rules and AGENTS.md; it also applies CLAUDE.md, for Claude Code compatibility.[4]
Skills, subagents and hooks
Skills in .agents/skills (the Agent Skills standard), custom agents and subagents, and hooks.[10]
Agent Skills, subagents in .cursor/agents and hooks; it also loads .claude and .codex skills and agents.[11]
MCP servers
STDIO and Streamable HTTP MCP servers in config.toml, from the app, CLI and IDE extension.[12]
Stdio, SSE and Streamable HTTP servers through mcp.json; Cloud Agents can use team MCP servers.[13]
Team analytics, built in
Workspace analytics of messages, credits and active users; a Codex dashboard and Analytics API on Enterprise.[14][15]
Teams dashboard: AI share of committed code, agent edits, messages by model, daily users and a leaderboard.[16]
Checking your own usage
A usage dashboard with limits and reset times; /status and /usage daily, weekly or cumulative in the CLI.[2][17]
A usage dashboard and editor settings show included usage per pool, Cursor models and other models.[3]
Each tool's column is read from its vendor's own docs and pricing page, dated in the sources below.
Codex or Cursor: which one fits your team
Pick Codex if your team is on ChatGPT: Plus, Pro and Business include it on the web, in the CLI, an IDE extension and the desktop app, with cloud tasks and @codex code review on GitHub. It runs OpenAI's GPT models. Heavy use is metered per five hours and per week, and past the limit you buy credits or switch to an API key.
Pick Cursor if the editor is where your engineers want the agent, with models from several vendors and Cursor's own Composer. A seat includes usage in two pools, Cursor's own models and everyone else's, and past both, on-demand usage continues at API prices.
They aren't either-or. Codex's IDE extension runs in Cursor, so an engineer can keep Cursor as the editor and hand a task to Codex from it.
Codex vs Cursor pricing and limits
Codex and Cursor, from each vendor's pricing page in October 2026:
Codex comes with ChatGPT Plus at $20 a month, Pro from $100 and Business at $25 per user ($20 billed yearly), and Free and Go ($8) run GPT-6 Luna in the desktop app. On Plus and Standard Business, OpenAI estimates 5–45 local messages per five hours on GPT-6 Astra, 15–160 on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna. More on Codex pricing.
Cursor Pro is $20 a month, Pro+ $60 and Ultra $200, and a Teams seat is $40 per user ($32 billed yearly). Cursor publishes no number for any plan's included usage: every plan includes a set amount, and the models you pick decide how fast it runs out. More on Cursor pricing.
What our own team saw
Codex wrote 31 of the 866 agent-written pull requests merged in our two repos from August 12 to October 4, 2026, all in September, with a median 576 lines changed, and they merged a median 37 minutes after opening. Cursor wrote none; it co-authored 7 of the 3,192 commits in our two repos from July 5 to October 3, 2026.
Running Codex and Cursor on one repository
Keep shared instructions in AGENTS.md: Codex reads it from your home folder down to the working directory, and Cursor reads it next to .cursor/rules and team rules. Skills in the SKILL.md format load in both (Cursor also loads Codex's), so one written for one tool works in the other.
TRY IT ON YOUR TEAM
See what your team ships with AI.
Starter is free (3 seats, no credit card). Pro is $49 a month with 5 seats and Max is $149 a month with 10 seats — priced per workspace, not per engineer, with extra seats at $19 a month.
Codex runs OpenAI's GPT models; its local clients can point at a custom Responses-API provider such as Ollama. Cursor offers models from OpenAI, Anthropic, Google and others.
Winding / Teams
