Skip to content
RuntaRunta
Start free with $50 credit

Architecture

Runta is an execution layer for AI agents. It provides scalable, governed runtimes with strong control over state, access, credentials, and execution.

Runta architecture with the entrance connected to a runtime platform containing runtimes surrounded by Secret stubs, Checkpoints, Token Xray, and Auto-scaling, plus network access control connecting to external APIs, repositories, and web resources.

Features

Section titled “Features”
Cloud AgentsCreate managed agents, send prompts, stream replies, access their computers, and manage their workspaces.
Runtime BasicsCreate, execute commands in, resize, pause, resume, stop, and delete runtimes.
Token X-RayIdentify potential token wasting patterns
Memory Auto ScalingAutomatically scale your agent’s resources based on usage.
FilesCopy, upload, download, read, and write files in a runtime.
EgressControl which external services your agent can access
CheckpointsCapture runtime state and restore it into new runtimes.
Secret StubsAuthenticate your agent’s requests without exposing credentials
Publish a ServiceMake your agent public accessible
Auto Suspend and Wake-UpSuspend your idle agents and wake them when needed
Team Access and RolesManage members from Settings → Team access with Owner, Admin, and Developer roles.

Reference

Section titled “Reference”
HarborRun Harbor jobs on the built-in Runta environment.
OpenAI Agents SDKRun OpenAI Agents SDK sandboxes on Runta runtimes.
CLI ReferenceCommands, flags, configuration, lifecycle, secrets, egress, checkpoints, and TLS.
SDK ReferencePython and TypeScript interfaces for runtimes, files, egress, secrets, and checkpoints.
REST API ReferenceGenerated API docs for runtimes, exec, checkpoints, files, egress, and secrets.
PricingUsage-based pricing for compute, memory, and storage.