Reliable
Every step is checked before the next
Independent Judges read the screen and confirm a step worked before Sai moves on. That is how Sai reached 79.38% on OSWorld 2.1, the highest partial score on the official leaderboard.
Compare
vs
Simular API vs Codex Computer Use
Codex Computer Use is a feature of the ChatGPT and Codex desktop apps: it operates the apps on your own Mac or Windows PC while you work in Codex, and it has no API. The Simular API is a computer-use agent you call from code or from any coding agent, on a hosted cloud computer or your own machine.
- Highest partial score on OSWorld 2.1
- 79.38%
- per task, the only cost published on the leaderboard
- $14.34
- to add Sai to Claude Code, Codex or Cursor
- 1 command
Simular API vs Codex Computer Use, side by side
As of October 2026
| Capability | Simular API | Codex Computer Use |
|---|---|---|
| What you get | ||
| Product | A working computer-use agent (Sai) behind one API | A feature (plugin) inside the ChatGPT / Codex desktop app, not an API |
| Computer included | Hosted Windows cloud computer, or your own Mac or Windows with the Sai desktop app | No cloud computer: runs on your own machine (or a host you connect) |
| Operating systems | Windows in the cloud; macOS and Windows on your own machine. No Linux | macOS (background) and Windows (foreground only). No Linux |
| Native desktop apps | Yes: desktop apps on Windows and macOS, not just a browser | Yes: any GUI app except terminals and ChatGPT itself |
| Agent loop and action execution | Built in: send a task in plain English, get the result | Built in, but only inside Codex; not callable from your own code |
| Setup / time to first task | Create an API key, add one MCP command (about a minute) | Low: install the app and plugin, grant Screen Recording and Accessibility |
| Reliability and cost | ||
| Benchmark | Highest partial score on the official OSWorld 2.1 leaderboard: 79.38% (Sai w/ Claude Opus 5, 500 steps, Oct 2026). Binary success 41.67%, #2 behind Claude Opus 5 (44.33%). Trajectories public on Hugging Face | No feature-level score. Model: GPT-6 Astra 72.6% on OSWorld 2.0 (Sep 22, 2026) |
| Step verification | Independent Judges check the screen before each next step | Not documented |
| Pricing | Credits per request. Free plan: one shared Windows cloud computer with daily limits. Paid: dedicated computers and top-ups. | Paid ChatGPT plan from Plus ($20/mo); regional limits |
| Cost per task / cost drivers | $14.34 per OSWorld 2.1 task, the only leaderboard entry with a published cost (OSWorld 2.0: $15.70) | Shares the Codex/ChatGPT 5-hour usage allowance plus credits |
| Repeat runs | Learns a workflow once, then replays it as self-healing code: up to 90% fewer tokens on repetitive tasks | Automations (scheduling) and memory; no action cache or replay found |
| Control and safety | ||
| Human approval | Built in: approve, approve for this task, or deny; link-only approvals for sign-ins and credentials | Per-app approval with 'Always allow', plus admin allowlists |
| Live view / takeover | Live view to watch tasks and answer approvals. No manual takeover | You see your own screen (PiP preview on macOS); stop or take over anytime |
| Data and compliance | SOC 2 Type II and HIPAA. Runs on Simular's cloud computer or your own machine. | Runs locally; screenshots processed by ChatGPT under your data controls |
| Operations | ||
| Sessions and persistence | Context carries across tasks on a computer; reset with new-session. | Your own desktop state persists; max session length not documented |
| Concurrency and limits | One task per computer at a time; 60 messages per hour; paid plans add dedicated computers | Several agents in parallel on one Mac; effectively one on Windows. Tied to one user's machine |
| Fit in your stack | ||
| Coding agents / MCP | Official MCP server (@simular-ai/sai-mcp) for Claude Code, Codex, Cursor and any stdio MCP client | Codex only; cannot be driven from Claude Code or Cursor |
| Model choice | Model-agnostic: Anthropic, OpenAI, Google, MiniMax, xAI, OpenRouter, or auto | OpenAI models only (GPT-6 Astra recommended) |
| SDKs / languages | HTTP API (curl, Python, Node examples) plus the MCP server on npm | Desktop app only (triggered with @Computer); no API or SDK support documented |
| Free tier | Yes: free plan includes a shared Windows cloud computer (daily limits) | No (Free/Go plans have no Computer Use) |
What unifies the Simular API
Four things every task gets, without extra code.
Efficient
Learn once, repeat for less
The first run works the task out. After that, Sai replays it as self-healing code instead of reasoning through every screen, using up to 90% fewer tokens on repetitive tasks.
Ready
A whole computer, already running
Sai works on a hosted Windows cloud computer, or on your own Mac or Windows machine. It opens apps, clicks and types in desktop software that has no API. No VMs or display drivers to manage.
Composable
Plugs into the agents you already use
Add Sai to Claude Code, Codex or Cursor with one MCP command, or call one HTTP endpoint from any language. It pauses for your approval before sign-ins and other sensitive steps.
Frequently asked questions
What is the difference between the Simular API and Codex Computer Use?
Codex Computer Use is a plugin in the ChatGPT and Codex desktop apps that sees and operates GUI apps on your own Mac or Windows computer while you work in Codex. The Simular API is an agent you call over HTTP or through an MCP server, on a hosted Windows cloud computer or on your own Mac or Windows machine.
Is there an API for Codex Computer Use?
No. It is triggered inside the desktop app (with @Computer), and no API or SDK support is documented. The Simular API is one HTTP endpoint, plus an official MCP server for Claude Code, Codex, Cursor and other MCP clients.
Can I use the Simular API from Codex?
Yes. Add the official MCP server (@simular-ai/sai-mcp) to Codex with one command, and Codex can hand computer tasks to Sai on a cloud computer instead of taking over your own screen.
Does Codex Computer Use run in the background on Windows?
No. On Windows it works in the foreground and takes over the pointer and keyboard of the session you are using. On macOS it can run in the background. Sai runs on its own Windows cloud computer, so your machine stays free.
Can Codex Computer Use control the terminal?
No. OpenAI’s documentation says it can’t automate terminal apps or ChatGPT itself. It can operate other GUI apps you approve, with an “Always allow” option and admin allowlists.
Is Codex Computer Use available in the EU and UK?
Not yet. OpenAI says initial availability excludes the European Union, Switzerland and the United Kingdom.
When is Codex Computer Use the better choice?
When you work in Codex on your own computer and want it to use your real apps and existing logins with no VM, run in the background on macOS, and be covered by flat plan pricing, phone remote control and admin app allowlists. Choose the Simular API when computer use has to run from your own code, product or any coding agent.
How do I get started with the Simular API?
Sign in at platform.simular.ai and create an API key. Add Sai to Claude Code, Codex or Cursor with one MCP command, or call the HTTP API from any language. The free plan includes a shared Windows cloud computer with daily limits.
Skip the harness. Ship the agent.
Get an API key in a minute, or talk to us about running Sai at scale.