Home / AI agent sandboxes / Anthropic sandbox-runtime

Anthropic sandbox-runtime

Agent sandbox

Open-source CLI and library that wraps any process in OS-level filesystem and network restrictions, no container needed.

Verdict (2026-11-13): Anthropic sandbox-runtime scores 46/100, #60 of 63 AI agent sandboxes. Strong on free tier, docs and SDKs. Weak on isolation, persistence, limits. Best for: Limiting what a local coding agent, MCP server or shell command can read, write and contact.

Skip it if: Shares the host kernel; weaker than VM-based isolation for hostile code.

Prices checked 2026-11-15.

List price per month
Self-host · no list price for 2 vCPU, 4 GiB, 176 hours a month (8 h × 22 days), CPU busy 30%
Free tier
Free open-source software under the Apache-2.0 license.
Isolation
OS-level process sandbox (sandbox-exec/Seatbelt on macOS, bubblewrap on Linux, restricted user account on Windows) with proxy-based network filtering
Pause / snapshots
Not stated / Not stated

sandbox-runtime (srt) restricts what a process and its children can read, write and reach on the network, using the operating system's own sandboxing: Seatbelt on macOS, bubblewrap on Linux and a dedicated user account with a Windows Filtering Platform fence on Windows. Network traffic goes through local proxies that enforce domain allow and deny lists. It was built for Claude Code and is published as an npm package that works as a CLI or a library. It is a research preview under Apache-2.0.

Best for

  • Limiting what a local coding agent, MCP server or shell command can read, write and contact
  • Adding domain allow lists to processes on a developer machine without running containers
  • Embedding OS-level sandboxing in a TypeScript tool

Watch out for

  • Shares the host kernel; weaker than VM-based isolation for hostile code
  • Beta research preview; APIs and configuration may change
  • Behavior and limits differ between macOS, Linux and Windows

Score breakdown

46/100 in AI agent sandboxes, rank 60 of 63. Parts with no data are left out and the rest is rescaled.

Price15/30Self-host, no list price: given the category median
Free tier10/10Free every month
Isolation4/15OS-level process sandbox
Persistence0/10No pause or snapshots stated
Limits0/10session limit not stated
Docs and SDKs9/102 SDKs or interfaces, open source
External benchmarks8.4/15Speed not measured: given the median of the 10 measured providers

Pricing

Storage
No separate rate published
Billing
Not applicable: free software run on your own machine

Full Anthropic sandbox-runtime pricing: plans and costs at four workloads →

Where Anthropic sandbox-runtime sits

406080100$1.06$9.40$83List price per month → (log scale)← cheaperAether: 61/100, $75.00Agent37 Cloud: 70/100, $1.06Alibaba Cloud Agent Sandbox: 68/100, $12.57Archil: 74/100, $33.44AWS Bedrock AgentCore Code Interpreter: 60/100, $16.10AWS Lambda MicroVMs: 65/100, $44.39Beam: 45/100, $24.64Blaxel: 63/100, $29.15boat: 71/100, $20.00Cloudflare Dynamic Workers: 48/100, $12.00Cloudflare Sandbox SDK: 56/100, $18.26CodeSandbox: 69/100, $26.40CreateOS Sandbox: 72/100, $20.89Daytona: 64/100, $29.15Depot: 55/100, $83.36DigitalOcean Agent Harness Runtime: 61/100, $50.00E2B: 68/100, $29.15Fly.io Sprites: 69/100, $19.47Freestyle: 61/100, $50.00Hopx: 56/100, $29.15Hyperbrowser: 62/100, $30.00InstaBox: 75/100, $20.00Isorun: 78/100, $19.36kosa8 Cloud: 58/100, $27.74Koyeb Sandboxes: 76/100, $10.14LangSmith Sandboxes: 72/100, $33.26Lizard Sandboxes: 76/100, $19.00Maritime: 59/100, $20.00MIOSA: 75/100, $20.00Modal Sandboxes: 57/100, $41.88Morph Cloud: 67/100, $17.60Mosaic Sandbox: 74/100, $8.80Namespace Devboxes: 62/100, $42.24Northflank: 78/100, $11.73Novita Agent Sandbox: 58/100, $20.53Podflare: 71/100, $28.86Promigence: 55/100, $29.15Railway Sandboxes: 59/100, $56.22Render Sandboxes: 63/100, $26.05Runloop: 62/100, $55.76Superserve: 69/100, $29.15Tenki Sandbox: 67/100, $29.15Tensorlake Sandboxes: 76/100, $17.95Upstash Box: 73/100, $10.56Vercel Sandbox: 76/100, $28.44Workpaper: 67/100, $15.00IsorunVercel SandboxMosaic SandboxCreateOS SandboxAgent37 Cloud
Score against list price per month for 2 vCPU, 4 GiB, 176 hours a month (8 h × 22 days), CPU busy 30%. Top left is better; a $0 price sits at the left edge. 46 of 63 providers publish a comparable price.

Anthropic sandbox-runtime alternatives

All 62 alternatives ranked →

Agent sandbox · #1

Isorun

78

$19.36 a month*

  • microVM
  • Free credit
  • Pause and resume
▲ 0
Agent sandbox · #2

Northflank

78

$11.73 a month*

  • microVM
  • Free monthly
  • Pause and resume
▲ 0
Agent sandbox · #3

Koyeb Sandboxes

76

$10.14 a month*

  • microVM
  • Pause and resume
  • Snapshots
▲ 0
Agent sandbox · #4

Lizard Sandboxes

76

$19.00 a month*

  • microVM
  • Free credit
  • Pause and resume
▲ 0
Agent sandbox · #5

Tensorlake Sandboxes

76

$17.95 a month*

  • microVM
  • Free monthly
  • Pause and resume
▲ 0
Agent sandbox · #6

Vercel Sandbox

76

$28.44 a month*

  • microVM
  • Free monthly
  • Pause and resume
▲ 0
Agent sandbox · #7

InstaBox

75

$20.00 a month*

  • microVM
  • Free monthly
  • Pause and resume
▲ 0
Agent sandbox · #8

MIOSA

75

$20.00 a month*

  • microVM
  • Free monthly
  • Pause and resume
▲ 0

Public benchmarks

Anthropic sandbox-runtime is not in the public benchmarks we track yet. See all benchmarks.

All facts

Anthropic sandbox-runtime
Categoryagent sandbox
Our score46/100
IsolationOS-level process sandbox (sandbox-exec/Seatbelt on macOS, bubblewrap on Linux, restricted user account on Windows) with proxy-based network filtering
BillingNot applicable: free software run on your own machine
Free tierFree open-source software under the Apache-2.0 license.
Network controlsYes
Open sourceYes (Apache-2.0)
Self-hostYes
SDKs and interfacesCLI, TypeScript/JavaScript library
RegionsRuns locally on macOS, Linux or Windows.
Facts checked2026-11-13

User reviews

No reviews yet. Have you used Anthropic sandbox-runtime? Your review helps other developers choose.

Sign in to review or upvote

Questions

Is Anthropic sandbox-runtime good?

Anthropic sandbox-runtime scores 46/100, #60 of 63 AI agent sandboxes. Strong on free tier, docs and SDKs. Weak on isolation, persistence, limits. Best for: Limiting what a local coding agent, MCP server or shell command can read, write and contact. Facts checked 2026-11-13.

How much does Anthropic sandbox-runtime cost?

Free open-source software under the Apache-2.0 license.

How is Anthropic sandbox-runtime isolated?

OS-level process sandbox (sandbox-exec/Seatbelt on macOS, bubblewrap on Linux, restricted user account on Windows) with proxy-based network filtering.

What are the best Anthropic sandbox-runtime alternatives?

In AI agent sandboxes, the top-rated are Isorun (78/100), Northflank (78/100), Koyeb Sandboxes (76/100). Anthropic sandbox-runtime is #60 of 63.

Sources

CompareTick 2 to 4 providers.