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 Claude Computer Use
Claude Computer Use gives a Claude model the tools to click and type; you build the computer, the agent loop and the safety around it. The Simular API gives you the finished agent on a ready computer. Send a task in plain English and get the result back.
- 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 Claude Computer Use, side by side
As of October 2026
| Capability | Simular API | Claude Computer Use |
|---|---|---|
| What you get | ||
| Product | A working computer-use agent (Sai) behind one API | A model toolset (computer_toolset_20260801, 17 actions). Not a hosted agent |
| Computer included | Hosted Windows cloud computer, or your own Mac or Windows with the Sai desktop app | No. You host the VM, container or desktop; Anthropic ships a reference Docker image |
| Operating systems | Windows in the cloud; macOS and Windows on your own machine. No Linux | Any desktop you can screenshot and drive; reference setup is Linux |
| Native desktop apps | Yes: desktop apps on Windows and macOS, not just a browser | Yes, in whatever environment you provide |
| Agent loop and action execution | Built in: send a task in plain English, get the result | No. You write the loop and the code that executes each action |
| Setup / time to first task | Create an API key, add one MCP command (about a minute) | Build the VM, action handlers, loop, coordinate scaling and safety. Demo runs with one docker run |
| 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 | OSWorld 2.1: Opus 5.5 81.8%, Opus 5 74.0% (labelled 'partial', Sep 22, 2026) |
| Step verification | Independent Judges check the screen before each next step | Not built in; up to your loop |
| Pricing | Credits per request. Free plan: one shared Windows cloud computer with daily limits. Paid: dedicated computers and top-ups. | Token prices only (e.g. Opus 5.5 $4/$20, Sonnet 5.5 $2/$10 per 1M). VM hosting is your cost |
| 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) | ~4,500 input tokens of toolset per request, plus 1,000–1,800 tokens per screenshot |
| Repeat runs | Learns a workflow once, then replays it as self-healing code: up to 90% fewer tokens on repetitive tasks | Prompt caching only; no workflow record/replay |
| Control and safety | ||
| Human approval | Built in: approve, approve for this task, or deny; link-only approvals for sign-ins and credentials | Recommended by Anthropic; you build it. Prompt-injection classifiers on by default |
| Live view / takeover | Live view to watch tasks and answer approvals. No manual takeover | None in the API; reference demo has VNC / noVNC viewer with interaction |
| Data and compliance | SOC 2 Type II and HIPAA. Runs on Simular's cloud computer or your own machine. | Screenshots and files stay in your environment; ZDR-eligible; SOC 2 Type II, ISO 27001 |
| Operations | ||
| Sessions and persistence | Context carries across tasks on a computer; reset with new-session. | Not applicable at API level; your VM decides persistence and duration |
| Concurrency and limits | One task per computer at a time; 60 messages per hour; paid plans add dedicated computers | API rate limits (e.g. Opus 5.5 start tier 1,000 RPM) plus however many VMs you run |
| Fit in your stack | ||
| Coding agents / MCP | Official MCP server (@simular-ai/sai-mcp) for Claude Code, Codex, Cursor and any stdio MCP client | Built-in computer-use MCP inside Claude Code (research preview: macOS CLI, Pro/Max; Desktop app macOS/Windows). Not for Codex or Cursor |
| Model choice | Model-agnostic: Anthropic, OpenAI, Google, MiniMax, xAI, OpenRouter, or auto | Claude models only |
| SDKs / languages | HTTP API (curl, Python, Node examples) plus the MCP server on npm | Python, TypeScript, C#, Go, Java, PHP, Ruby; also on Bedrock, Vertex, Foundry |
| Free tier | Yes: free plan includes a shared Windows cloud computer (daily limits) | No API free tier; pay as you go (new orgs start in an evaluation tier) |
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 Claude Computer Use?
Claude Computer Use is a toolset (computer_toolset_20260801, 17 actions such as screenshot, click and type) that lets a Claude model request actions. Your application runs the computer, executes each action and loops the results back. The Simular API is a complete computer-use agent: you send a task in plain English, Sai does the work on a ready computer and returns the result. If you want a working agent rather than a toolkit, it is a Claude Computer Use alternative.
How much does Claude Computer Use cost compared with the Simular API?
Claude Computer Use is billed in tokens: the toolset adds about 4,500 input tokens to every request, each screenshot adds 1,000–1,800 more, and the computer you run it on is your cost. The Simular API is billed in credits per request. On OSWorld 2.1 Sai completed tasks at $14.34 each, the only cost published on the leaderboard, and on repetitive tasks it uses up to 90% fewer tokens by replaying learned workflows as code.
Do I need to set up a virtual machine for Claude Computer Use?
Yes. Anthropic recommends a dedicated virtual machine or container with minimal privileges and a virtual display, and ships a Linux Docker reference implementation. With the Simular API the computer is included: a hosted Windows cloud computer, or your own Mac or Windows machine with the Sai desktop app.
Can I still use Claude models with the Simular API?
Yes. The Simular API is model-agnostic: Anthropic, OpenAI, Google, MiniMax, xAI and OpenRouter models, or auto. Sai’s OSWorld 2.1 result was run with Claude Opus 5.
Can I use computer use inside Claude Code?
Claude Code has a built-in computer-use MCP server as a research preview: macOS in the CLI, on Pro and Max plans, and in the Claude Desktop app on macOS and Windows. Sai’s official MCP server (@simular-ai/sai-mcp) works in Claude Code, Codex, Cursor and any stdio MCP client, and runs tasks on a cloud computer instead of your own screen.
Which is more accurate on OSWorld 2.1?
On the official OSWorld 2.1 leaderboard (500 steps), Sai with Claude Opus 5 has the highest partial score, 79.38%; its binary success rate is 41.67%, second to Claude Opus 5 at 44.33%. Anthropic self-reports 81.8% for Claude Opus 5.5 on OSWorld 2.1, but that model is not on the official leaderboard.
When is Claude Computer Use the better choice?
When you want full data control and zero data retention, need any environment including air-gapped ones, prefer token-only pricing, or need availability across clouds. Choose the Simular API when you want a working agent today without building and maintaining the computer, loop, step checks and approvals yourself.
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.