Sandboxes

Trusted infrastructure for untrusted agents

High performance sandboxes designed to give agents the computers they need at a scale they demand.

Computers for agents.

Reinforcement learning rollouts

import modalapp = modal.App("code-rl-reward")@app.function()def compute_reward(completion: str, testcase: list[str]) -> int:    code = get_generated_code_and_test_cases(completion, testcase)    # Each rollout gets its own isolated, disposable sandbox    sb = modal.Sandbox.create(app=app)    score = 0    try:        p = sb.exec("python", "-c", code, timeout=30)        p.wait()        score = 1 if p.returncode == 0 else 0   # reward = did the code run?    finally:        sb.terminate()    return score

SANDBOXES

1M+ in parallel

Rollouts that scale fast without slowing down your workflow.

Get started

1M+ sandboxes in under a minute

Keep your GPUs saturated and rollouts fast with massively parallel sandboxes that spin up fast.
Any dependency, any image

Programmatically define your exact rollout environment and dependencies.
Egress controls to prevent cheating

Lock each rollout's egress, keeping rollout environments reproducible and contained.

Coding Agents

import modalapp = modal.App.lookup("coding-agent", create_if_missing=True)sandbox = modal.Sandbox.create(app=app)try:    code = agent.generate_code("Calculate the first 20 Fibonacci numbers")    result = sandbox.exec("python", "-c", code, timeout=30)    agent.process_result(result)finally:    sandbox.terminate()

SANDBOX

Started

Environment

agent.py - tools.py

Image

Resources

Volumes

Task

Executing untrusted code

python / workspace / agent.py

The fast, reliable computer your coding agent needs to do its best work.

Get started

Fast startup on any image

Snapshot full or partial filesystems to save state and resume instantly later.
Secure networking

Lock down domain egress while still giving agents access to the Secrets, data, and tools they need.
Build robust environments

Start from any image, install what you need at runtime, and keep state across a long agent session.

import modalapp = modal.App("code-rl-reward")@app.function()def compute_reward(completion: str, testcase: list[str]) -> int:    code = get_generated_code_and_test_cases(completion, testcase)    # Each rollout gets its own isolated, disposable sandbox    sb = modal.Sandbox.create(app=app)    score = 0    try:        p = sb.exec("python", "-c", code, timeout=30)        p.wait()        score = 1 if p.returncode == 0 else 0   # reward = did the code run?    finally:        sb.terminate()    return score

SANDBOXES

1M+ in parallel

Rollouts that scale fast without slowing down your workflow.

Get started

1M+ sandboxes in under a minute

Keep your GPUs saturated and rollouts fast with massively parallel sandboxes that spin up fast.
Any dependency, any image

Programmatically define your exact rollout environment and dependencies.
Egress controls to prevent cheating

Lock each rollout's egress, keeping rollout environments reproducible and contained.

Engineered for the speed and scale agents require

Scale to a million in under a minute

Modal's infrastructure scales to the parallelism that production agent systems and RL training actually demand.

Learn more

Fast on any image

Bring your dependencies, your base image, your setup scripts. Sub-second scheduling and strong cold-start performance on custom images

Learn more

Adaptive provisioning of resources

Elastically scale both CPU and Memory as needed so you never have to worry about under provisioning or over paying.

Learn more

Modal Sandbox Pricing

Only pay for what you use. Burst up to what you need without over-allocating CPU or memory in advance.

Compute costs

CPU

Physical core (2 vCPU equivalent)

$0.1419 / core / h

*minimum of 0.125 cores per container

Memory

$0.0240 / GiB / h

GPU

See standard pricing

Reinforcement Learning

“Modal powers both our reinforcement learning infrastructure and production inference. Millions of sandboxes on one end, real-time serving on the other.”

Scott Wu CEO, Cognition

Coding Agents

“We use Sandboxes to power Coding Sessions, which allows users to delegate and iterate on issues with their coding agent by tagging @Linear.”

Ryan Delaney Software Engineer, Linear

Coding Agents

“It would have been really hard to build Inspect without Modal. A lot of the complexity is hidden behind just being able to spin up sandboxes very quickly with a lot of services and data, and effectively have infinite of them at any given time.”

Rahul Sengottuvelu Head of Applied AI, Ramp

Blog