SPRINT Shop Machine
Software and tool connections, preconfigured for each new Machine.
pnpm · agent-browser
AGENTS.mdRepository structure and development commandsticket-workflowTicket workflow and review requirementsTHE WORKSPACE FOR YOUR AGENTS
Thinair brings your agents, their tools, and their computers into one workspace. Describe an agent to build it, give it work, and open its applications to collaborate.
Mac and Cloud. Available by invitation.
Your agent has a computer.
You have the whole workspace.
See the work. Use the tools. Step in.
Your agent is handling the ticket.
Go and handle the ticket SPR-214.
I’ll read the ticket and find the right project setup.
function addToBag() {
bag.add({ product });
}01 / 03One instruction. Your agent goes to Linear.
Launch a task. Tell your agent what needs to get done.
Prepare your machine and agent once.
Give every task its own agent and computer.
Open any agent to talk, or use any browser directly. You and your agents share the same applications and controls.
MACHINES FOR YOUR AGENTSTHINAIR / 02
Thinair runs your agents on computers configured with your dependencies, repository, and credentials.
Define the software in a Machine Definition. Store credentials in a Profile. Reuse both on your Mac or in Thinair Cloud.
Tools in a Machine Definition. Subscription sign-ins and API keys in a Profile.
Software and tool connections, preconfigured for each new Machine.
pnpm · agent-browser
AGENTS.mdRepository structure and development commandsticket-workflowTicket workflow and review requirementsBring the software and credentials together. Choose where it runs and which agent to open.
SPR-214~/create/sprintThe preview shows your selected Machine Definition, Profile, repository, and location.
Save the Machine Disk, then launch from that state with the Profile you choose.
src/product.tsxtests/product-bag.spec.tsThe Snapshot preserves this Machine Disk at a point in time.
SPRINT Work Profile supplies the credentials when you launch.
Launch a Machine with these saved files and tools.
Snapshots capture the Machine Disk. Attached Drives and Profile credentials stay separate.
Machine Definition: SPRINT Shop Machine. Profile: SPRINT Work Profile. Repository: sprint / storefront. Agent: Codex, Default model. Location: A Linux VM on your Mac. Select Preview this Machine to see the result.
Choose how your agent thinks, what it can use, and where it works. Each part has a purpose. You put them together.
Thinair is harness-agnostic. Bring Codex, Claude Code, OpenCode, Aider, or your own agent. Give it a prepared Machine and keep the workflow you know.
Thinair is model-agnostic. Choose commercial or open models through your harness. Bring your own subscription, API keys, or a local model server.
Save the tools, dependencies, skills, and instructions your project needs. Start another Machine from the same prepared setup.
Bring your agent sign-ins, service credentials, and browser accounts. Keep identity separate from the software in your Machine Definition.
Run a Linux Machine on your Mac or in Thinair Cloud. Your agent uses its apps and files. You can inspect the work or take control.
Ask a question, describe a change, or explain what you want to create. Your agent starts with a clear task and a prepared computer.
WHERE IT RUNSTHINAIR / 04
Open Thinair in your browser and work on Cloud Machines. No installation required. Use Thinair Native to run Linux VMs on your Mac and connect to Cloud Machines.
Your hardware. Your Linux VMs.
~/launch-page main
❯ bun run dev
▲ Next.js
Local localhost:3000
✓ Ready in 842ms
GET / 200 38ms
GET /api/status 200
▍
Web development
Animate the product. Export an MP4.
The Blender scene is rendering. I’ll export the MP4 with FFmpeg.
Rendering frames
Product animation
2 Linux VMs on your Mac
Your agents and apps run on Cloud Machines.
Analyze this week’s sales.
I’ve queried PostgreSQL. I’m charting the sales in Jupyter.
Preparing report
Data analysis & reporting
~/automations main
❯ ./browser-checks.sh
✓ Docker services ready
✓ Chrome session open
✓ Checkout flow passed
✓ Screenshots saved
Browser checks complete.
Listening for the next job
Browser automation
2 Linux VMs in Thinair Cloud
Three parts: the software to install, the credentials to use, and the computer that runs it.
A reusable recipe that installs the apps, dependencies, and project code your work needs.
Your agent subscriptions, API keys, and browser sign-ins. Every app on the Machine uses its selected Profile.
An isolated Linux virtual machine with its own CPU, memory, and disk. It runs your agent, terminal, and installed software.
Capture the Machine Disk. Launch a new Machine from the Snapshot.
Use agents, terminals, and apps in your browser. No installation required.
Thinair Native runs Linux VMs on your Mac and connects to Cloud Machines. Choose where each task runs.
HOW YOU WORKTHINAIR / 05
Spread out in a Workspace. Focus in a Workbench. Bring your own terminal or agent through CLI and MCP. Your tools, ideas, and everything they make — together.
The page is running in Chrome. I’ve added an image, a reference video, and two Components to explore.
An experiment in light,
material, and motion.

orbit main ❯ pnpm dev
VITE ready in 184 ms
› Local: http://localhost:3000
Watching for file changes.
orbit-render.webp 1020 × 840
Talk, plan, and follow along. The agent works in the same tools you do.
Inspect Machines, call Thinair operations, and use the same tool catalog as the agent. Keep working from your own terminal.
Example output from an authenticated Thinair connection.~/work
❯thinair --json rpc machine list --input '{}'
[
{ "name": "studio-mac",
"platform": "linux" },
{ "name": "studio-cloud",
"platform": "linux" }
]
Excerpt · two Machines, two locations❯
Give a compatible agent access to Thinair tools over MCP. Use your preferred client to work with the same resources.
Illustration of an authenticated MCP connection. Available tools depend on access.Your agent discovers the tools
Thinair makes available to it.
ONE PLACE FOR THE WHOLE PROCESS.Illustrated workspace. Try the Components and switch views.
Thinair for Mac and Thinair Cloud are in Early Access, available by invitation. Enter the access code you received on the Early Access page to continue to the Mac download or Cloud.
Thinair provides the workspace and infrastructure for building and orchestrating AI agents. It brings together your agents, their tools, and the computers they run on.
No. All of Thinair’s systems are exposed through CLI, MCP, and API. Use your own agents, terminals, scripts, or applications to configure environments, launch Machines, and coordinate work. Use Thinair as infrastructure, through its interface, or both.
Describe what you want your agent to do. Thinair prepares its tools and environment, helps you connect the credentials it needs, and launches it in your workspace. Then give it a task and start working together.
Software development, 3D modeling, video production, research, data analysis, and marketing. Configure each agent with the software, skills, and credentials its work requires.
Give different agents different jobs, follow their progress, and coordinate their work from one workspace or through your own tools. You decide what each agent works on and what happens next.
Yes. Open their browsers, terminals, files, and applications yourself. Inspect results, make changes, and give feedback in the same environment where the agent is working.
Use Codex or Claude Code with a supported subscription or API key. Run other agents through Terminal and connect the models they support. Your provider’s usage limits still apply.
Thinair Web runs in your browser and works with Cloud Machines, with no installation required. Thinair Native is the Mac app: run Linux VMs on your Mac and connect to Cloud Machines.
Yes. Save its software and tools in a Machine Definition and its credentials in a Profile. Reuse them to launch another Machine, or return to an existing Machine to continue working.
YOUR NEXT AGENT
Give your agent a task and work alongside it. Mac and Cloud are in Early Access, available by invitation.
Use your own tools through CLI, MCP, and API.
Read the getting-started guide