Sai is an always-on agentic AI coworker that can actually do work on a computer for you. Sai can use apps, websites, and desktop interfaces, navigating the graphic user interface (GUI) like humans do. It combines the behavior of an AI agent, an AI personal assistant, and a workflow automation tool in one product. Our vision is to free humans from mundane computer work so they can get back to what matters.

Sai is for anyone stuck doing repetitive, time-consuming computer tasks. Business professionals, founders, researchers, and non-technical users alike use it as a personal AI assistant that does more than answer questions — it actually gets work done.

What can I use Sai for?

You can use Sai for a wide range of digital workflows, including research, lead sourcing, follow-up drafting, scheduling, file organization, spreadsheet updates, document handling, recurring reporting, and other repetitive work across apps and websites. Sai is especially useful for tasks that require real computer interaction.

How is Sai different from Claude Cowork, Perplexity Computer, OpenClaw, and other AI agents?

Sai is designed around four core differentiators: a secure workspace where the AI works, GUI-based computer use (not just browser or API actions), approval-based control for important actions, and always-on background execution so work can continue while you do other things. Sai is built to support broader digital work by combining browser, desktop GUI, terminal, and tool-based workflows in one system, while keeping users in control through live visibility and approvals.

What is a GUI-based agent?

A GUI-based agent is an AI agent that can interact with the graphical user interface of software, meaning it can work with buttons, menus, windows, text fields, and other visible elements on a computer screen. Instead of relying only on APIs or text commands, a GUI-based agent can operate software the way a human user does. Simular's agentic framework Agent S was the first to outperform human performance on the OSWorld benchmark.

What does BYOD mean in Sai?

BYOD stands for Bring Your Own Device. In Sai, BYOD means you can use your own device environment to access Sai's capabilities instead of relying only on the default setup. Sai's BYOD supports devices including Mac mini, macOS and Windows.

How do I start using Sai?

Sai is designed to help users get started quickly without complex setup. You can give Sai a task, choose a workflow, or describe the outcome you want — all in plain English, and Sai will work inside its secure workspace.

Does Sai support skills?

Yes. Sai supports skills, so you can extend how it works for different tasks and workflows. This makes it easier to adapt Sai to more specific use cases rather than relying only on one default behavior. It also supercharges collaboration with the most effective skill shared across a team.

Is Sai safe for business workflows?

Sai is built with production-grade, enterprise use in mind, with a dedicated digital workspace, approval flows, and live visibility that give teams the safety and oversight of real business demands.

Blog

Product updateJune 30, 2026

Introducing the Sai API: Put Your Agent to Work, Programmatically

On this page

Today we’re launching the Sai HTTP API — a way to send tasks to your Sai agent over HTTPS, straight from your own code. If you’ve ever wanted Sai to run inside a CI pipeline, on a device, or as part of your own app, this is for you.

It’s the same REST surface that powers our sapi CLI, now documented and available to build on directly.

Here’s a task from start to finish — one request in, a live stream out:

bash

curl -N -X POST https://api.sai.simular.ai/v1/agents/message \
  -H "Authorization: Bearer sapi_..." \
  -H "Content-Type: application/json" \
  -d '{"machineId":"m_1a2b","message":"list my unread emails from today"}'

text

data: {"type":"start"}
data: {"type":"reasoning-delta","id":"r1","delta":"Opening Gmail..."}
data: {"type":"text-delta","id":"t1","delta":"You have 3 unread emails: ..."}
data: {"type":"finish","finishReason":"stop"}
data: [DONE]

What you can do

Send tasks to your agent — Fire off a message like “list my unread emails from today” and stream the agent’s response back in real time.
Manage sessions and machines — List your machines, start fresh conversations, resume earlier ones, and pull recent context.
Attach files — Upload files to include with a message.
Stay in control — Approvals, aborts, and restarts are all available over the API, so the agent asks before doing anything sensitive — even when you call it programmatically.

Built for real workflows

Live streaming responses — The send-message endpoint returns a Server-Sent Events stream, so you get the agent’s narration, tool activity, and final answer as they happen.
Simple, secure auth — Authenticate with a long-lived API key (great for servers, CI, and devices) or an interactive sign-in. Generate keys with sapi key generate or from Settings → API Keys in the desktop app.
One base URL — Everything lives under https://api.sai.simular.ai, with agent endpoints under /v1/agents and account endpoints under /v1/account. Requests and responses are JSON.

bash

curl https://api.sai.simular.ai/v1/agents/auth \
  -H "Authorization: Bearer sapi_your_api_key_here"

Prefer the command line?

The same capabilities are available through the sapi CLI — perfect for quick scripts and terminal-driven workflows.

Get started

Read the full reference, including authentication, endpoints, streaming events, and rate limits, here:

Sai API Documentation

We can’t wait to see what you build with Sai.

Questions or feedback? Come say hi in our community — we’d love to hear how you’re using the API.

Email us at support@simular.ai.

Back to Blog