The coding agent is already wired up.
Run official headless harnesses with managed provider credentials. Your app sends tasks through the Platform API; Sandpi handles harness execution.
Codex · Claude Code · Pi · Kimi Code · ZCode
Managed agents / Sandpi Platform
Integrate Codex, Claude Code, and more into your product through Sandpi Platform. Harness execution, sandboxes, session history, and files are managed for you.
No additional sandbox charge.
Pay for model tokens. A separate 5% fee applies to credit top-ups.
You build
Your application
Sandpi Platform
Managed for youNative agent harnesses
Built on Sandbox0
Isolated, persistent workspaces
01Less infrastructure to own
An agent integration needs more than compute. It needs a harness, a workspace, conversation continuity, and a way to bring results back. Sandpi puts those pieces behind an API, so you can focus on what your agent does for your users.
Run official headless harnesses with managed provider credentials. Your app sends tasks through the Platform API; Sandpi handles harness execution.
Codex · Claude Code · Pi · Kimi Code · ZCode
Project files live in a persistent Environment. Sessions retain native conversation history across Runs, so follow-up tasks can continue the same work.
Persistent files + native session history
Stream and replay Run events, inspect file changes, read workspace files, and retrieve itemized model usage through the API.
Events + file changes + usage
Fork a configured Environment for a new task, stop an active Run, or roll back the latest Run once its checkpoint is ready.
Environment forks + Run checkpoints
02Built for integration
One API workflow across harnesses. Persistent Environments hold the project; Sessions hold native history; Runs do the work.
Choose your harness
Codex uses OpenAI. A Session keeps its harness and provider; each Run selects a supported model.
POST/v1/environments/{environmentId}/sessions
{
"harness": "codex"
}Authenticated with your Sandpi Platform API token.
Start a persistent project workspace. Wait until it is ready.
Choose a harness. Sandpi retains its native conversation history.
Send a task and a model. Stream events, read files, and inspect usage.
03Sandbox cost, removed
$0
Additional Platform sandbox charge
No separate sandbox bill for your Platform Runs. Model tokens are billed at official provider rates, with a separate 5% fee when you top up prepaid credits.
Prepaid credit is required to create or fork Environments and submit Runs. No personal Sandpi TUI subscription is required.
04Choose your layer
Sandbox infrastructure
Build and operate your own agent runtime with sandbox primitives and deployment control.
Managed agent harnesses
Bring coding agents into your product through an API, with the execution stack managed for you.
Before you integrate
Sandpi Platform runs coding-agent harnesses in persistent Sandbox0 Environments. You integrate tasks, Sessions, events, files, and usage through an API. Sandpi manages harness execution and the underlying sandbox lifecycle.
Sandbox usage for Sandpi Platform Runs has no additional Platform charge. Model tokens are billed at official provider rates, and prepaid credit top-ups carry a separate 5% fee. You need prepaid credit to create or fork an Environment or submit a Run. Sandpi's personal TUI subscription plans are separate and are not required for Platform.
A Session fixes its harness and provider. To use another harness, create another Session. Native history is retained within the same Session; it is not transparently transferred between harnesses. Only one Run mutates an Environment at a time.
Use Sandbox0 when you want to build your own agent runtime, run a custom harness, or control templates, processes, network policy, and self-hosting. Use Sandpi Platform when you want managed coding-agent harnesses integrated into your application through an API.
Built on Sandbox0. Delivered by Sandpi.
Start with a harness. Keep the workspace. Ship the integration.
Continue to sandpi.ai/platform