Bashkit
Awesomely fast virtual sandbox with bash and file system. Written in Rust.
Features
- Secure by default - No process spawning, no filesystem access, no network access unless explicitly enabled. 250+ threats analyzed and mitigated
- POSIX compliant - Substantial IEEE 1003.1-2024 Shell Command Language compliance
- Sandboxed, in-process execution - All 160 commands reimplemented in Rust, no
fork/exec - Virtual filesystem - InMemoryFs, OverlayFs, MountableFs with optional RealFs backend (
realfsfeature) - Resource limits - Command count, loop iterations, function depth, output size, filesystem size, parser fuel
- Network allowlist - HTTP access denied by default, per-domain control
- Multi-tenant isolation - Each interpreter instance is fully independent
- Custom builtins - Extend with domain-specific commands
- LLM tool contract -
BashToolwith discovery metadata, streaming output, and system prompts - Snapshotting - Serialize shell state and VFS contents for checkpoint/resume workflows
- Scripted tool orchestration - Compose ToolDef+callback pairs into multi-tool bash scripts (
scripted_toolfeature) - MCP server - Model Context Protocol endpoint via
bashkit mcp - Async-first - Built on tokio
- Language bindings - Python (PyO3) and JavaScript/TypeScript (NAPI-RS) for Node.js, Bun, and Deno
- Experimental: Git support - Virtual git operations on the virtual filesystem (
gitfeature) - Experimental: Python support - Embedded Python interpreter via Monty (
pythonfeature) - Experimental: TypeScript support - Embedded TypeScript interpreter via ZapCode (
typescriptfeature)
Install
Or add to Cargo.toml:
[]
= "0.1"
Optional features:
Quick Start
use Bash;
async
LLM Tool Contract
BashTool follows the toolkit-library contract: builder for reusable config,
immutable tool metadata for discovery, and single-use executions for each call.
use ;
use StreamExt;
#
# async
Overview
Built-in Commands (160)
| Category | Commands |
|---|---|
| Core | echo, printf, cat, nl, read, mapfile, readarray |
| Navigation | cd, pwd, ls, tree, find, pushd, popd, dirs |
| Flow control | true, false, exit, return, break, continue, test, [ |
| Variables | export, set, unset, local, shift, source, ., eval, readonly, times, declare, typeset, let, alias, unalias |
| Shell | bash, sh (virtual re-invocation), :, trap, caller, getopts, shopt, command, type, which, hash, compgen, fc, help |
| Text processing | grep, rg, sed, awk, jq, head, tail, sort, uniq, cut, tr, wc, paste, column, diff, comm, strings, tac, rev, seq, expr, fold, expand, unexpand, join, iconv |
| File operations | mkdir, mktemp, mkfifo, rm, cp, mv, touch, chmod, chown, ln, rmdir, realpath, readlink, split |
| File inspection | file, stat, less |
| Archives | tar, gzip, gunzip, zip, unzip |
| Byte tools | od, xxd, hexdump, base64 |
| Checksums | md5sum, sha1sum, sha256sum |
| Utilities | sleep, date, basename, dirname, timeout, wait, watch, yes, kill, bc, clear |
| Disk | df, du |
| Pipeline | xargs, tee |
| System info | whoami, hostname, uname, id, env, printenv, history |
| Data formats | csv, json, yaml, tomlq, template, envsubst |
| Network | curl, wget (requires allowlist), http |
| DevOps | assert, dotenv, glob, log, retry, semver, verify, parallel, patch |
| Experimental | python, python3 (requires python feature), ts, typescript, node, deno, bun (requires typescript feature), git (requires git feature) |
Shell Features
- Variables and parameter expansion (
$VAR,${VAR:-default},${#VAR},${var@Q}, case conversion${var^^}) - Command substitution (
$(cmd),`cmd`) - Arithmetic expansion (
$((1 + 2)),declare -i,let) - Pipelines and redirections (
|,>,>>,<,<<<,2>&1,&>) - Control flow (
if/elif/else,for,while,until,casewith;;/;&/;;&,select) - Functions (POSIX and bash-style) with dynamic scoping, FUNCNAME stack,
caller - Indexed arrays (
arr=(a b c),${arr[@]},${#arr[@]}, slicing,+=) - Associative arrays (
declare -A map=([key]=val)) - Nameref variables (
declare -n) - Brace expansion (
{a,b,c},{1..10},{01..05}) - Glob expansion (
*,?) and extended globs (@(),?(),*(),+(),!()) - Glob options (
dotglob,nullglob,failglob,nocaseglob,globstar) - Here documents (
<<EOF,<<-EOFwith tab stripping,<<<here-strings) - Process substitution (
<(cmd),>(cmd)) - Coprocesses (
coproc) - Background execution (
&) withwait - Shell options (
set -euxo pipefail,shopt) - Alias expansion
- Trap handling (
trap cmd EXIT,trap cmd ERR) [[ ]]conditionals with regex matching (=~, BASH_REMATCH)
Configuration
use ;
use Arc;
let limits = new
.max_commands
.max_loop_iterations
.max_function_depth;
let mut bash = builder
.fs
.env
.cwd
.limits
.build;
Virtual Identity
Configure the virtual username and hostname for whoami, hostname, id, and uname:
let mut bash = builder
.username // Sets whoami, id, and $USER env var
.hostname // Sets hostname, uname -n
.build;
// whoami → "deploy"
// hostname → "my-server"
// id → "uid=1000(deploy) gid=1000(deploy)..."
// echo $USER → "deploy"
Snapshotting
Checkpoint an interpreter to bytes, then restore it later:
use ;
#
# async
See docs/snapshotting.md for Rust, Python, and Node examples, plus snapshot security notes.
Experimental: Git Support
Enable the git feature for virtual git operations on the virtual filesystem.
All git data lives in the VFS — no host filesystem access.
[]
= { = "0.1", = ["git"] }
use ;
let mut bash = builder
.git
.build;
// Local operations: init, add, commit, status, log
// Branch operations: branch, checkout, diff, reset
// Remote operations: remote add/remove, clone/push/pull/fetch (virtual mode)
See specs/git-support.md for the full specification.
Experimental: Python Support
Enable the python feature to embed the Monty Python interpreter (pure Rust, Python 3.12).
Python code runs in-memory with configurable resource limits and VFS bridging — files created
by bash are readable from Python and vice versa.
[]
= { = "0.1", = ["python"] }
use Bash;
let mut bash = builder.python.build;
// Inline code
bash.exec.await?;
// Script files from VFS
bash.exec.await?;
// VFS bridging: pathlib.Path operations work with the virtual filesystem
bash.exec.await?;
bash.exec.await?; // "hello from python"
Stdlib modules: math, pathlib, os (getenv/environ), sys, typing.
Security note: re is intentionally disabled due to regex backtracking DoS risk.
Limitations: no open() (use pathlib.Path), no network, no classes, no third-party imports.
See crates/bashkit/docs/python.md for the full guide.
Experimental: TypeScript Support
Enable the typescript feature to embed the ZapCode TypeScript interpreter (pure Rust, no V8).
TypeScript code runs in-memory with configurable resource limits and VFS bridging via external function suspend/resume.
[]
= { = "0.1", = ["typescript"] }
use Bash;
let mut bash = builder.typescript.build;
// Inline code (ts, node, deno, bun aliases all work)
bash.exec.await?;
bash.exec.await?;
// Script files from VFS
bash.exec.await?;
// VFS bridging: readFile/writeFile async functions
bash.exec.await?;
bash.exec.await?; // "hello from ts"
Compat aliases (node, deno, bun) and unsupported-mode hints are configurable:
use ;
// Only ts/typescript, no compat aliases
let bash = builder
.typescript_with_config
.build;
Limitations: no import/require, no eval(), no network, no process/Deno/Bun globals.
See crates/bashkit/docs/typescript.md for the full guide.
Virtual Filesystem
use ;
use Arc;
// Layer filesystems
let base = new;
let overlay = new;
// Mount points
let mut mountable = new;
mountable.mount;
CLI Usage
# Run a script
# Interactive REPL
# MCP server (Model Context Protocol)
# Mount real filesystem (read-only or read-write)
Development
LLM Eval Results
Bashkit includes an eval harness that measures how well LLMs use bashkit as a bash tool in agentic workloads — 58 tasks across 15 categories.
| Model | Score | Tasks Passed | Tool Call Success | Duration |
|---|---|---|---|---|
| Claude Haiku 4.5 | 97% | 54/58 | 88% | 8.6 min |
| Claude Sonnet 4.6 | 93% | 48/58 | 85% | 20.5 min |
| Claude Opus 4.6 | 91% | 50/58 | 88% | 20.1 min |
| GPT-5.3-Codex | 91% | 51/58 | 83% | 19.6 min |
| GPT-5.2 | 77% | 41/58 | 67% | 7.0 min |
Delta from v0.1.7 (on shared 37 tasks): Haiku 98%→100%, Opus 93%→96%, GPT-5.2 86%→86% (3 more tasks). Interpreter fixes unblocked json_to_csv_export and script_function_lib across models. See the detailed analysis.
Benchmarks
Bashkit includes a benchmark tool to compare performance against bash and just-bash.
See crates/bashkit-bench/README.md for methodology and assumptions.
Language Bindings
Python
Python bindings with LangChain integration are available in crates/bashkit-python.
=
= await
JavaScript / TypeScript
NAPI-RS bindings for Node.js, Bun, and Deno. Available as @everruns/bashkit on npm.
import { BashTool } from '@everruns/bashkit';
const tool = new BashTool({ username: 'agent', hostname: 'sandbox' });
const result = await tool.execute("echo 'Hello, World!'");
console.log(result.stdout);
// Direct VFS access
await tool.writeFile('/tmp/data.txt', 'hello');
const content = await tool.readFile('/tmp/data.txt');
Platform matrix: macOS (x86_64, aarch64), Linux (x86_64, aarch64), Windows (x86_64), WASM. See crates/bashkit-js for details.
Security
Bashkit is built for running untrusted scripts from AI agents and users. Security is a core design goal, not an afterthought.
Defense in Depth
| Layer | Protection |
|---|---|
| No process spawning | All 160 commands are reimplemented in Rust — no fork, exec, or shell escape |
| Virtual filesystem | Scripts see an in-memory FS by default; no host filesystem access unless explicitly mounted |
| Network allowlist | HTTP access is denied by default; each domain must be explicitly allowed |
| Resource limits | Configurable caps on commands (10K), loop iterations (100K), function depth (100), output (10MB), input (10MB) |
| Filesystem limits | Max total bytes (100MB), max file size (10MB), max file count (10K) — prevents zip bombs, tar bombs, and append floods |
| Parser limits | Timeout (5s), fuel budget (100K ops), AST depth (100) — prevents pathological input from hanging the interpreter |
| Multi-tenant isolation | Each Bash instance is fully isolated — no shared state between tenants |
| Panic recovery | All builtins wrapped in catch_unwind — a panic in one command doesn't crash the host |
| Path traversal prevention | RealFs backend canonicalizes paths to prevent ../../etc/passwd escapes |
| Unicode security | 68 byte-boundary tests across builtins; zero-width character rejection in VFS paths |
Threat Model
60+ identified threats across 11 categories (DoS, sandbox escape, info disclosure, injection, network, isolation, internal errors, git, logging, Python, Unicode) — each with a stable ID, mitigation status, and test coverage.
See the threat model for the full analysis and security policy for reporting vulnerabilities.
Other Virtual Bash Implementations
- just-bash (TypeScript, Apache-2.0) — Virtual bash interpreter for AI agents by Vercel Labs. Custom recursive descent parser, 75+ reimplemented commands (including full awk/sed/jq), in-memory VFS, defense-in-depth sandboxing, AST transform plugins. Runs in Node.js and browser.
- gbash (Go, Apache-2.0) — Deterministic, sandbox-only bash runtime for AI agents. Delegates parsing to
mvdan/sh. Registry-backed commands, policy enforcement, structured tracing, JSON-RPC server mode.
Acknowledgments
Bashkit is an independent implementation that draws design inspiration from several open source projects:
- just-bash (Vercel Labs, Apache-2.0) — Pioneered the idea of a virtual bash interpreter for AI-powered environments. Bashkit's sandboxing architecture and multi-tenant design was inspired by their approach.
- Oils (Andy Chu, Apache-2.0) — Comprehensive bash compatibility testing approach inspired our spec test methodology.
- One True AWK (Lucent Technologies) — AWK language semantics reference for our awk builtin.
- jq (Stephen Dolan, MIT) — jq query syntax and behavior reference. Our implementation uses the jaq Rust crates.
No code was copied from any of these projects. See NOTICE for full details.
Contributing
The best way to contribute is to open an issue — bug reports, feature requests, and questions all help improve bashkit. If you'd like to contribute code, see CONTRIBUTING.md for setup and workflow details.
Ecosystem
Bashkit is part of the Everruns ecosystem.
License
MIT