- Own workspace
- Own browser profile
- Client accounts
Sessions · separate conversation histories
A persistent home for your agents
A workspace, a browser, and the accounts your agent needs, together in one Environment. Leave a task running and return from another device.
Sessions · separate conversation histories
Sessions · separate conversation histories
Isolated between Environments · Shared within each Environment
The Environment is the computer
Each cloud Environment owns one isolated Sandbox and a persistent workspace. Give a project its own place to run, with its own browser profile and agent credentials.
Files, repositories, installed tools, and native agent history stay in the Environment across reconnects and pause/resume.
Each cloud Environment keeps its own browser profile and website logins. You and your agent use the same browser inside it.
Configure agent credentials and network access per Environment. Separate cloud Environments keep project runtime and state apart.
The Session is the task history
Start a Session for a feature, another for research, and another for maintenance. Each keeps its own native conversation history while using the same Environment's workspace, tools, browser logins, and Sandbox.
Sessions share resources and can change the same files. Separate history does not create a separate isolation boundary.
ONE CLOUD ENVIRONMENT
Separate histories. Shared working state and compute.
One browser. The same working state.
Open a website and complete its normal login.
Use the same tabs and login to work on the task.
Tabs and logins are shared inside this Environment.
A real browser for the work
Open a website, complete its login yourself, and let your agent continue with the same tabs and authenticated state. Inspect an app running in the Sandbox, gather research, or review a report together.
Browser profile data persists in the Environment. Different cloud Environments can keep different website accounts; Sessions within one Environment share that profile.
Native agents, lasting work
Run Codex, Claude Code, Pi, Kimi Code, or ZCode through its native terminal. Keep commands and tools familiar while Sandpi manages the Environment around them.
Reconnect to the terminal from a desktop, phone, or tablet. A controller takeover lets another device continue the same work.
Save workspace snapshots, restore them, or fork a project into another Environment when you need a separate working copy.
Schedule recurring work or let an external event trigger an agent task in its established Environment.
Put it to work
Put client work and personal work in different cloud Environments, each with its own workspace, browser profile, and agent credentials.
Sign in to a website yourself, then let your agent use the same browser to research, inspect an app, or prepare a report.
Leave a task in the cloud. Check in from your phone, then return to its terminal, files, and native history on your laptop.
Or connect your machine
Connect your computer and run its installed agent in a selected local workspace. Your machine's availability and OS account permissions apply. Use cloud Environments when you want Sandbox0 isolation and a task that stays available without keeping your computer online.
Choose your workflow
Both use persistent Sandbox0 Environments. Each product has its own Environments, Sessions, and billing.
| Your workflow | Sandpi · sandpi.ai | Sandpi Platform |
|---|---|---|
| How you work | Control a native agent terminal, files, and browser yourself. | Your service creates Sessions, submits Runs, and streams events through an API. |
| Agent access | Sign in to your agent or supply its credentials. | Bring your model API key and an endpoint compatible with the native harness. |
| What you pay for | A hosted Environment plan. Your provider bills model usage separately. | Shared Sandbox runtime, billed by the second. Your model provider or gateway bills model usage directly. |
| Where it runs | A cloud Environment, or your own connected machine. | A persistent cloud Environment for headless work. |
Before you start
Sandpi supports the native Codex, Claude Code, Pi, Kimi Code, and ZCode terminal experiences. Each Environment has a selected agent; use that agent's own commands, tools, and conversation history.
Sessions have separate conversation histories, but share their Environment's Sandbox, files, tools, browser profile, and credentials. Changes made in one Session can affect another. Use a separate cloud Environment when you need independent state or accounts.
No. A new Session in an existing cloud Environment reuses its Sandbox. You can keep several task histories without adding a hosted Environment for each conversation. Sessions share the Environment's available compute; this does not add capacity or guarantee simultaneous execution.
Yes. Each cloud Environment has its own agent credentials and persistent browser profile. You can sign in to websites with different accounts in different Environments. Sessions inside the same Environment share those logins.
Yes. Connect your machine and select a local workspace. It uses your installed agent and OS account permissions, and the machine must remain available. Cloud Environments provide Sandbox0 isolation; connecting a personal machine does not create that isolation on your computer.
Sandpi is for interacting with an agent through its native terminal and browser. Platform is for software that submits headless Runs through an API, using your model API key and compatible endpoint. The two products have separate Environments, Sessions, and billing.
Ready when the work is