@anthropic-ai/sandbox-runtime
Advanced tools
Anthropic Sandbox Runtime (ASRT) - A general-purpose tool for wrapping security boundaries around arbitrary processes
A lightweight sandboxing tool for enforcing filesystem and network restrictions on arbitrary processes at the OS level, without requiring a container.
srt uses native OS sandboxing primitives (sandbox-exec on macOS, bubblewrap on Linux) and proxy-based network filtering. It can be used to sandbox the behaviour of agents, local MCP servers, bash commands and arbitrary processes.
Beta Research Preview
The Sandbox Runtime is a research preview developed for Claude Code to enable safer AI agents. It's being made available as an early open source preview to help the broader ecosystem build more secure agentic systems. As this is an early research preview, APIs and configuration formats may evolve. We welcome feedback and contributions to make AI agents safer by default!
npm install @anthropic-ai/sandbox-runtime
# Network restrictions
$ srt "curl anthropic.com"
Running: curl anthropic.com
<html>...</html> # Request succeeds
$ srt "curl example.com"
Running: curl example.com
Connection blocked by network allowlist # Request blocked
# Filesystem restrictions
$ srt "cat README.md"
Running: cat README.md
# Anthropic Sandb... # Current directory access allowed
$ srt "cat ~/.ssh/id_rsa"
Running: cat ~/.ssh/id_rsa
cat: /Users/ollie/.ssh/id_rsa: Operation not permitted # Specific file blocked
This package provides a standalone sandbox implementation that can be used as both a CLI tool and a library. It's designed with a secure-by-default philosophy tailored for common developer use cases: processes start with minimal access, and you explicitly poke only the holes you need.
Key capabilities:
A key use case is sandboxing Model Context Protocol (MCP) servers to restrict their capabilities. For example, to sandbox the filesystem MCP server:
Without sandboxing (.mcp.json):
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem"]
}
}
}
With sandboxing (.mcp.json):
{
"mcpServers": {
"filesystem": {
"command": "srt",
"args": ["npx", "-y", "@modelcontextprotocol/server-filesystem"]
}
}
}
Then configure restrictions in ~/.claude/settings.json:
{
"permissions": {
"deny": [
"Edit(~/sensitive-folder)"
]
}
}
Now the MCP server will be blocked from writing to the denied path:
> Write a file to ~/sensitive-folder
✗ Error: EPERM: operation not permitted, open '/Users/ollie/sensitive-folder/test.txt'
The sandbox uses OS-level primitives to enforce restrictions that apply to the entire process tree:
sandbox-exec with dynamically generated Seatbelt profilesBoth filesystem and network isolation are required for effective sandboxing. Without network isolation, a compromised process could exfiltrate SSH keys or other sensitive files. Without filesystem isolation, a process could escape the sandbox and gain unrestricted network access.
Filesystem Isolation enforces read and write restrictions:
Read(~/.ssh))Edit(/tmp)), and deny within allowed paths (e.g., deny Edit(./secret) even though . is allowed)Network Isolation routes all traffic through proxy servers running on the host:
Linux: Requests are routed via the filesystem over a Unix domain socket. The network namespace of the sandboxed process is removed entirely, so all network traffic must go through the proxies running on the host (listening on Unix sockets that are bind-mounted into the sandbox)
macOS: The Seatbelt profile allows communication only to a specific localhost port. The proxies listen on this port, creating a controlled channel for all network access
Both HTTP/HTTPS (via HTTP proxy) and other TCP traffic (via SOCKS5 proxy) are mediated by these proxies, which enforce your domain allowlists and denylists.
For more details on sandboxing in Claude Code, see:
src/
├── index.ts # Library exports
├── cli.ts # CLI entrypoint (srt command)
├── utils/ # Shared utilities
│ ├── debug.ts # Debug logging
│ ├── settings.ts # Settings reader (permissions + sandbox config)
│ ├── platform.ts # Platform detection
│ └── exec.ts # Command execution utilities
└── sandbox/ # Sandbox implementation
├── sandbox-manager.ts # Main sandbox manager
├── sandbox-schemas.ts # Zod schemas for validation
├── sandbox-violation-store.ts # Violation tracking
├── sandbox-utils.ts # Shared sandbox utilities
├── http-proxy.ts # HTTP/HTTPS proxy for network filtering
├── socks-proxy.ts # SOCKS5 proxy for network filtering
├── linux-sandbox-utils.ts # Linux bubblewrap sandboxing
└── macos-sandbox-utils.ts # macOS sandbox-exec sandboxing
The srt command (Anthropic Sandbox Runtime) wraps any command with security boundaries:
# Run a command in the sandbox
srt echo "hello world"
# With debug logging
srt --debug curl https://example.com
# Specify custom settings file
srt --settings /path/to/settings.json npm install
import { SandboxManager } from '@anthropic-ai/sandbox-runtime'
import { spawn } from 'child_process'
// Initialize the sandbox (starts proxy servers, etc.)
await SandboxManager.initialize()
// Wrap a command with sandbox restrictions
const sandboxedCommand = await SandboxManager.wrapWithSandbox('curl https://example.com')
// Execute the sandboxed command
const child = spawn(sandboxedCommand, { shell: true, stdio: 'inherit' })
// Handle exit
child.on('exit', (code) => {
console.log(`Command exited with code ${code}`)
})
// Cleanup when done (optional, happens automatically on process exit)
await SandboxManager.reset()
// Main sandbox manager
export { SandboxManager } from '@anthropic-ai/sandbox-runtime'
// Violation tracking
export { SandboxViolationStore } from '@anthropic-ai/sandbox-runtime'
// TypeScript types
export type {
SandboxAskCallback,
IgnoreViolationsConfig,
FsReadRestrictionConfig,
FsWriteRestrictionConfig,
NetworkRestrictionConfig,
} from '@anthropic-ai/sandbox-runtime'
NOTE: Configuration format is provisional and subject to change. The current configuration system is borrowed from Claude Code's permission model and uses Claude Code settings files. This format may evolve as the sandbox runtime matures as a standalone tool.
Settings are loaded and merged in priority order from:
~/.claude/settings.json$CWD/.claude/settings.json$CWD/.claude/settings.local.json/Library/Application Support/ClaudeCode/managed-settings.json/etc/claude-code/managed-settings.json--settings flag{
"sandbox": {
"enabled": true,
"network": {
"allowUnixSockets": ["/var/run/docker.sock"],
"allowLocalBinding": false,
"httpProxyPort": 8888,
"socksProxyPort": 1080
}
},
"permissions": {
"allow": [
"WebFetch(domain:github.com)",
"WebFetch(domain:lfs.github.com)",
"WebFetch(domain:api.github.com)",
"WebFetch(domain:npmjs.org)",
"Edit(src/)",
"Edit(test/)",
"Read(.)"
],
"deny": [
"Edit(.env)",
"Edit(config/production.json)",
"Read(~/.ssh)",
"WebFetch(domain:malicious.com)"
]
}
}
sandbox.enabled - Enable/disable sandboxing (boolean)sandbox.network.allowUnixSockets - Unix sockets that can be accessed (array of paths)sandbox.network.allowLocalBinding - Allow binding to local ports (boolean)sandbox.network.httpProxyPort - Port for your own HTTP proxy (default: uses built-in proxy)sandbox.network.socksProxyPort - Port for your own SOCKS5 proxy (default: uses built-in proxy)Permission rules provide fine-grained control and use Claude Code's permission syntax:
WebFetch(domain:example.com) - Allow/deny network access to a domainEdit(path) - Allow/deny file write accessRead(path) - Allow/deny file read accessPath Syntax (macOS):
Paths support git-style glob patterns on macOS, similar to .gitignore syntax:
* - Matches any characters except / (e.g., *.ts matches foo.ts but not foo/bar.ts)** - Matches any characters including / (e.g., src/**/*.ts matches all .ts files in src/)? - Matches any single character except / (e.g., file?.txt matches file1.txt)[abc] - Matches any character in the set (e.g., file[0-9].txt matches file3.txt)Examples:
Edit(src/) - Allow write to entire src/ directoryEdit(src/**/*.ts) - Allow write to all .ts files in src/ and subdirectoriesRead(~/.ssh) - Deny read to SSH directoryEdit(.env) - Deny write to .env file (even if current directory is allowed)Path Syntax (Linux):
Linux currently does not support glob matching. Use literal paths only:
Edit(src/) - Allow write to src/ directoryRead(/home/user/.ssh) - Deny read to SSH directoryAll platforms:
/home/user/.ssh) or relative to the current working directory (e.g., ./src)~ expands to the user's home directoryAllow GitHub access (all necessary endpoints):
{
"permissions": {
"allow": [
"WebFetch(domain:github.com)",
"WebFetch(domain:lfs.github.com)",
"WebFetch(domain:api.github.com)"
]
}
}
Restrict to specific directories:
{
"permissions": {
"allow": [
"Edit(src/)",
"Edit(test/)",
"Read(.)"
],
"deny": [
"Edit(.env)",
"Edit(secrets/)",
"Read(~/.ssh)"
]
}
}
Running Jest: Use --no-watchman flag to avoid sandbox violations:
srt "jest --no-watchman"
Watchman accesses files outside the sandbox boundaries, which will trigger permission errors. Disabling it allows Jest to run with the built-in file watcher instead.
sandbox-exec with custom profiles (no additional dependencies)bubblewrap (bwrap) for containerizationLinux requires:
bubblewrap - Container runtime
apt-get install bubblewrapdnf install bubblewrappacman -S bubblewrapsocat - Socket relay for proxy bridging
apt-get install socatdnf install socatpacman -S socatripgrep - Fast search tool for deny path detection
apt-get install ripgrepdnf install ripgreppacman -S ripgrepmacOS requires:
ripgrep - Fast search tool for deny path detection
brew install ripgrep# Install dependencies
npm install
# Build the project
npm run build
# Type checking
npm run typecheck
# Lint code
npm run lint
# Format code
npm run format
The sandbox runs HTTP and SOCKS5 proxy servers on the host machine that filter all network requests based on permission rules:
permissions rules from your configurationPlatform-specific proxy communication:
Linux: Requests are routed via the filesystem over Unix domain sockets (using socat for bridging). The network namespace is removed from the bubblewrap container, ensuring all network traffic must go through the proxies.
macOS: The Seatbelt profile allows communication only to specific localhost ports where the proxies listen. All other network access is blocked.
Filesystem restrictions are enforced at the OS level:
sandbox-exec with dynamically generated Seatbelt profiles that specify allowed read/write pathsbubblewrap with bind mounts, marking directories as read-only or read-write based on configurationDefault filesystem permissions:
Read: Allowed everywhere by default. You can deny specific paths using Read(path) deny rules
Read(~/.ssh) to block access to SSH keysWrite: Only allowed in the current working directory by default. You can:
Edit(path) allow rules (e.g., Edit(/tmp))Edit(.env) even though . is allowed)This model lets you start with broad read access but tightly controlled write access, then refine both as needed.
When a sandboxed process attempts to access a restricted resource:
EPERM error)macOS: The sandbox runtime taps into macOS's system sandbox violation log store. This provides real-time notifications with detailed information about what was attempted and why it was blocked. This is the same mechanism Claude Code uses for violation detection.
# View sandbox violations in real-time
log stream --predicate 'process == "sandbox-exec"' --style syslog
Linux: Bubblewrap doesn't provide built-in violation reporting. Use strace to trace system calls and identify blocked operations:
# Trace all denied operations
strace -f srt <your-command> 2>&1 | grep EPERM
# Trace specific file operations
strace -f -e trace=open,openat,stat,access srt <your-command> 2>&1 | grep EPERM
# Trace network operations
strace -f -e trace=network srt <your-command> 2>&1 | grep EPERM
For more sophisticated network filtering, you can configure the sandbox to use your own proxy instead of the built-in ones. This enables:
Example with mitmproxy:
# Start mitmproxy with custom filtering script
mitmproxy -s custom_filter.py --listen-port 8888
# Configure sandbox to use your proxy
{
"sandbox": {
"network": {
"httpProxyPort": 8888
}
}
}
Important security consideration: Even with domain allowlists, exfiltration vectors may exist. For example, allowing github.com lets a process push to any repository. With a custom MITM proxy and proper certificate setup, you can inspect and filter specific API calls to prevent this.
allowUnixSockets configuration can inadvertently grant access to powerful system services that could lead to sandbox bypasses. For example, if it is used to allow access to /var/run/docker.sock this would effectively grant access to the host system through exploiting the docker socket. Users are encouraged to carefully consider any unix sockets that they allow through the sandbox.$PATH, system configuration directories, or user shell configuration files (.bashrc, .zshrc) can lead to code execution in different security contexts when other users or system processes access these files.enableWeakerNestedSandbox mode that enables it to work inside of Docker environments without privileged namespaces. This option considerably weakens security and should only be used incases where additional isolation is otherwise enforced.Linux proxy bypass: Currently uses environment variables (HTTP_PROXY, HTTPS_PROXY, ALL_PROXY) to direct traffic through proxies. This works for most applications but may be ignored by programs that don't respect these variables, leading to them being unable to connect to the internet.
Future improvements:
Proxychains support: Add support for proxychains with LD_PRELOAD on Linux to intercept network calls at a lower level, making bypass more difficult
Linux violation monitoring: Implement automatic strace-based violation detection for Linux, integrated with the violation store. Currently, Linux users must manually run strace to see violations, unlike macOS which has automatic violation monitoring via the system log store
FAQs
Anthropic Sandbox Runtime (ASRT) - A general-purpose tool for wrapping security boundaries around arbitrary processes
The npm package @anthropic-ai/sandbox-runtime receives a total of 0 weekly downloads. As such, @anthropic-ai/sandbox-runtime popularity was classified as not popular.
We found that @anthropic-ai/sandbox-runtime demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 15 open source maintainers collaborating on the project.
Threat Research
Malicious Firefox extensions mimic Phantom, Jupiter, and Bitget wallets to steal recovery phrases and private keys, with one fetching its payload remotely.
Product
Socket Threat Intel now shows which malware campaigns were blocked or detected in your organization, with links to scans and Firewall events.
Security News
Fixes for eMudhra and Brotli are out. Getting them everywhere they’re needed means tracking down bundled libraries, runtimes, and certificate stores.