LIVE
Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|Google's AI gives evasive answers about who won the 2020 US election28/10/26 · Google|NVIDIA introduces Nemotron 3.5 Transcribe, a promptable multilingual transcription model28/10/26 · NVIDIA|OpenAI and ServiceNow deepen partnership on enterprise workflows28/10/26 · OpenAI|Failed startups' data finds a second life training AI models28/10/26|Anthropic expands AI for Science program to 20 foundations and research institutions28/10/26 · Anthropic|Anthropic shares early results on using Claude to accelerate drug development27/10/26 · Anthropic|GEPA-Curator: Filtering Failures to Better Train LLM Agents27/10/26|Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|Google's AI gives evasive answers about who won the 2020 US election28/10/26 · Google|NVIDIA introduces Nemotron 3.5 Transcribe, a promptable multilingual transcription model28/10/26 · NVIDIA|OpenAI and ServiceNow deepen partnership on enterprise workflows28/10/26 · OpenAI|Failed startups' data finds a second life training AI models28/10/26|Anthropic expands AI for Science program to 20 foundations and research institutions28/10/26 · Anthropic|Anthropic shares early results on using Claude to accelerate drug development27/10/26 · Anthropic|GEPA-Curator: Filtering Failures to Better Train LLM Agents27/10/26|
Toolofficial publisher

Bash (Claude Code)

Tool : A precise action the AI can call · requires claude-code

Executes shell commands in a persistent session.

Essential and dangerous: use pattern-based allowlists, never --dangerously-skip-permissions outside a container.

Overall score
7.1 /10
Security: serious reservations
Utility
9.0
Reliability
6.0
Security
Serious reservations
Ease
5 min · Comfortable

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

Bash is the built-in Claude Code tool that executes shell commands directly in your environment within a persistent session. It addresses the need to run scripts, builds, tests, or any command-line operation without leaving the working context.

In practice, Claude Code submits a command to your machine's shell and retrieves the output to incorporate it into the conversation. The session is persistent: environment variables, the current directory, and launched processes remain active across commands. A set of read-only commands runs without asking for your approval; all others trigger a permission prompt in Manual mode.

This tool is aimed at developers who want to delegate automation, build, or command-line debugging tasks to Claude. It suits those who know their environment well enough to define precise permission rules.

It is not suitable for non-technical users or contexts where running arbitrary commands on the host machine poses an unacceptable risk without prior isolation.

What it offers

  • ·Executes shell commands in a persistent session that retains environment variables and the current directory between calls.
  • ·Applies permission rules by command pattern using allow, deny, and ask formats in permission settings.
  • ·Runs a built-in set of read-only commands without prompting, even in Manual mode.
  • ·Integrates with PreToolUse and PostToolUse hooks to intercept or log each call before or after execution.
  • ·Accepts rules such as Bash(npm run *) to allow or block specific families of commands.
  • ·Can be restricted or extended via --allowedTools and --disallowedTools options in the CLI or Agent SDK.

In practice

Price
Free
Commercial use
Commercial use: yes
Available on
claude-code
Host
claude-code
FR interface
No

Security and data

Rating
Security: serious reservations
Capabilities
readswritesexecutesnetwork
Your data
Data stays local
Maintenance
Active

Links and repository checked on 25 October 2026 · added on 1 February 2025

How to install Bash (Claude Code)

1. Install Claude Code using the standard procedure (npm or equivalent). 2. Bash is included by default: no separate installation is needed. 3. Open your Claude Code settings file and add your rules under permissions.allow and permissions.deny using the Bash(pattern) format. 4. For sensitive commands, use the ask format so that confirmation is requested each time. 5. If you are working inside an isolated container, you may relax restrictions; outside a container, keep a strict pattern-based allowlist.
permissions.allow: ["Bash(git *)","Bash(npm test)"]

Official documentation ↗

In the same category

Whole category Code →