Coding agents

Claude Code

Give Claude Code a computer by adding the Sai MCP server.

The @simular-ai/sai-mcp server exposes Sai as MCP tools. Claude Code starts it over stdio and calls it like any other tool.

Install

InstallShell

claude mcp add sai -e SAI_API_KEY=sapi_... -- npx -y @simular-ai/sai-mcp

Add -s user to make it available in every project. Requires Node.js 22.12 or newer.

Verify

VerifyShell

claude mcp list

sai should be listed as connected. Then, in a Claude Code session:

Prompt

Use Sai to open Notepad, type hello, and save it to the desktop as hello.txt.

Environment

Variable	Required	Notes
`SAI_API_KEY`	yes	An `sapi_` key from platform.simular.ai.
`SAI_API_URL`	no	Defaults to production. Set it to point at staging.
`SAI_PUSH`	no	Only for channels. See Channels.
`SAI_LOCAL_VIEW`	no	`0` turns off the local `watch_url` page. See Watch Sai work.
`SAI_CONTROL`	no	`0` leaves out the control tools, which are on by default. See Direct control.

Tools

Tool	Input	Output
`sai_machines`	none	`{ machines: [{ machineId, name, kind, online }], default }`
`sai_models`	none	`{ models: [{ id, name, provider, tier, costTier, allowed }], default }`
`sai_task_start`	`{ task, machine_id?, file_ids?, new_session?, model?, autonomous? }`	`{ session_id, machine_id, queued, push, task_id, watch_url? }`
`sai_task_wait`	`{ session_id, timeout_s?, cursor? }`	`{ status, events, text, approval?, usage?, sessionUsage?, cursor }`
`sai_task_approve`	`{ approval_id, decision, selections? }`	`{ ok }`, or `{ ok: false, approval_url, reason: "link_only" }`
`sai_task_abort`	`{ session_id }`	`{ ok }`
`sai_upload`	`{ path }`	`{ file_id, name, mime, size }`
`sai_live_link`	`{ machine_id? }`	`{ machine_id, live_url, expires_at }`

usage and sessionUsage are the task's and the session's model tokens and cost so far, counted live and not the billing figure; see Token usage. decision is approve, approve_for_task or deny. timeout_s defaults to 45 and is capped at 55, under the 60-second tool timeout most MCP clients use. model is an id from sai_models or auto; it sticks to the session until changed, and the model only calls sai_models when you ask for a specific model.

sai_live_link returns a link to the computer's screen that a person can open without a Simular sign-in: live_url, valid until expires_at (Unix time in milliseconds). The mouse and keyboard work there; the page pauses its own input while a Sai task is running, but that is not enforced. machine_id is optional when the account has one computer. Anyone with the link can use the screen until it expires, so keep it to yourself. See When Sai needs a person.

autonomous: true on sai_task_start (from @simular-ai/sai-mcp 0.3.0) runs that task without permission requests, as if approve_for_task had been answered up front; the grant ends with the task. The agent only sets it when you ask for a task to run unattended. Approvals that need a person in a browser, such as a sign-in, still stop the task.

What the model does

The server's instructions describe the loop, so you do not have to prompt for it:

sai_task_start with the task.
sai_task_wait until status is not running.
On needs_approval: sai_task_approve when the approval can be answered, otherwise show approval_url to you and wait again.
On idle: report text.
When Sai says it needs you on the screen (a sign-in, a captcha, a code): sai_live_link, and give you live_url. Open it in any browser, do the step, and tell the agent; it sends Sai a message with sai_task_start to carry on.

There is no screenshot tool. Sai describes what it sees in text and progress events. To watch the screen yourself, open the watch_url link the agent gives you: see Watch Sai work.

Gotchas

One conversation per computer. Pass new_session: true to start clean.
Files go through sai_upload first; then pass the returned file_id in file_ids.
A free account's first task of the day can sit in running for about two minutes while the computer starts.
If sai_task_wait times out, call it again with the returned cursor. The task keeps running.
sai_task_wait retries a gateway 502, 503 or 504 on its own, three times with a growing pause, before reporting an error. Other statuses come back at once.

Optional: the sai skill

The package also ships a Claude Code skill that teaches the loop, and falls back to plain HTTP in a session where the MCP server is not loaded. Install it with:

Install the skillShell

npx -y @simular-ai/sai-mcp init-claude            # for you: ~/.claude/skills/sai
npx -y @simular-ai/sai-mcp init-claude --project  # for this repo: ./.claude/skills/sai

Re-run it after upgrading to refresh the skill. It never overwrites a SKILL.md you edited unless you pass --force.