Loading...
Loading...
Agent Sandbox
Sandboxed environments with configurable isolation, resource limits, and full execution logging for every AI agent action.
Capabilities
Complete process isolation with no network or filesystem access. Agents execute in fully sandboxed environments.
Set CPU, memory, and time limits per sandbox. Prevent runaway processes from consuming shared infrastructure.
Every command executed is logged with full input/output capture, exit codes, and resource usage metrics.
Set expiration dates on sandboxes. Expired environments are automatically cleaned up with no manual intervention.
Choose from full, network-only, filesystem-only, or minimal isolation depending on your security requirements.
Sandboxes are scoped to workspaces with agent-level assignment. Only authorized agents can access their environments.
Getting Started
Define a sandbox with isolation level, resource limits, and optional agent assignment. Choose from four isolation modes.
Run commands inside the sandbox with full input/output capture. Monitor resource usage and execution duration in real time.
Review execution logs, resource usage, and outcomes. Terminate sandboxes when no longer needed or let them auto-expire.
Developer Integration
import { DRD } from '@drd/sdk';
const drd = new DRD({ token: 'drd_live_sk_...' });
// Create an isolated sandbox
const sandbox = await drd.sandbox.create({
name: 'data-processor',
isolationLevel: 'full',
config: { maxMemoryMB: 512, timeoutMs: 30000 },
});
// Execute a command
const result = await drd.sandbox.execute({
sandboxId: sandbox.id,
command: 'process_data',
input: { source: 's3://bucket/data.csv' },
});
console.log(result.output, result.durationMs);Isolate AI agent execution with resource controls, full logging, and automatic expiration.