my-claude-code

skill
Guvenlik Denetimi
Basarisiz
Health Uyari
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 9 GitHub stars
Code Basarisiz
  • network request — Outbound network request in package.json
  • spawnSync — Synchronous process spawning in packages/@ant/computer-use-input/src/index.ts
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This tool is a personal fork of the official Anthropic Claude Code CLI. It is modified to remove telemetry, bypass safety prompt restrictions, and add multi-provider support (OpenAI, Bedrock, Vertex AI).

Security Assessment
Overall Risk: High. This fork fundamentally alters the security boundaries of the original CLI by reducing safety prompt layers, which could allow the underlying AI models to execute more dangerous commands without standard restrictions. The code makes outbound network requests to connect to various AI providers. Most critically, the tool uses synchronous process spawning (`spawnSync`), meaning it can execute commands directly on your local system. While no hardcoded secrets or dangerous permissions were found, piping a remote install script (`curl | bash`) directly into your shell is always a high-risk action, especially with modified CLI tools.

Quality Assessment
The project is actively maintained, with its most recent push occurring today. However, it suffers from low community visibility (only 9 GitHub stars), which means few independent developers have reviewed the code for vulnerabilities or malicious behavior. Additionally, the repository lacks a formal license file, meaning the legal terms for using, modifying, or distributing this modified version of Anthropic's software are entirely unclear.

Verdict
Use with caution (or Not recommended for enterprise environments) due to stripped safety guardrails, active local code execution capabilities, and unvetted origin.
SUMMARY

Personal build of Claude Code with multi-provider support (Anthropic, OpenAI Codex, Bedrock, Vertex AI)

README.md

my-claude-code

A personal build of Claude Code with multi-provider support and experimental feature flags.

version Bun TypeScript React Anthropic Claude OpenAI Codex AWS Bedrock Google Vertex AI

my-claude-code demo


Table of Contents


What This Repository Is

my-claude-code is a personal fork of Anthropic's Claude Code CLI, maintained as a public build workspace.

This fork makes three practical changes over upstream:

  • telemetry-related behavior is removed or stubbed where possible
  • prompt-layer restrictions added by the CLI wrapper are reduced
  • build-time experimental feature flags are exposed for local builds and testing

Quick Start

Option 1: Clone and build locally

git clone https://github.com/Icarus603/my-claude-code.git
cd my-claude-code
bun install
bun run build
./dist/cli.js

Option 2: One-line installer

curl -fsSL https://raw.githubusercontent.com/Icarus603/my-claude-code/main/install.sh | bash

The installer checks your system, installs Bun if needed, clones the repo, builds dist/cli.js, and symlinks my-claude-code into ~/.local/bin.

Uninstall

To completely remove my-claude-code from your system:

# Remove the binary symlink
rm ~/.local/bin/my-claude-code

# Remove the installation directory
rm -rf ~/my-claude-code

# Remove configuration and cache (optional)
rm -rf ~/.my-claude-code

Authentication and Providers

This repo supports multiple providers, but they do not all authenticate the same way.

Anthropic

Use Anthropic directly with either:

  • ANTHROPIC_API_KEY
  • /login for Anthropic OAuth
export ANTHROPIC_API_KEY="..."
./dist/cli.js

OpenAI Codex

Codex authenticates via OpenAI OAuth through the CLI:

./dist/cli.js

Inside the CLI, run /login and choose the Codex login flow.

After a successful Codex login, CLAUDE_CODE_USE_OPENAI=1 is persisted in your user config automatically, so new terminal sessions will continue using OpenAI Codex without a manual export.

Supported model examples:

Model ID
GPT-5.3 Codex gpt-5.3-codex
GPT-5.4 gpt-5.4
GPT-5.4 Mini gpt-5.4-mini

AWS Bedrock

Bedrock does not use /login. It uses AWS credentials plus provider flags.

export CLAUDE_CODE_USE_BEDROCK=1
export AWS_REGION="us-east-1"
./dist/cli.js

Google Vertex AI

Vertex does not use /login. It uses Google Cloud credentials.

gcloud auth application-default login
export CLAUDE_CODE_USE_VERTEX=1
./dist/cli.js

Anthropic Foundry

Foundry does not use /login. It uses explicit environment configuration.

export CLAUDE_CODE_USE_FOUNDRY=1
export ANTHROPIC_FOUNDRY_API_KEY="..."
./dist/cli.js

Provider Summary

Provider Selection Authentication
Anthropic default ANTHROPIC_API_KEY or Anthropic /login
OpenAI Codex default after Codex login OpenAI /login
AWS Bedrock CLAUDE_CODE_USE_BEDROCK=1 AWS credentials
Google Vertex AI CLAUDE_CODE_USE_VERTEX=1 Google ADC
Anthropic Foundry CLAUDE_CODE_USE_FOUNDRY=1 ANTHROPIC_FOUNDRY_API_KEY

Build and Run

Requirements

  • Bun >= 1.3.11
  • macOS or Linux
  • valid credentials for whichever provider you plan to use

Install Bun if needed:

curl -fsSL https://bun.sh/install | bash

Standard build

bun install
bun run build
./dist/cli.js

Build output

Command Output Notes
bun run build ./dist/cli.js compiled standalone binary
bun run dev source execution slower startup, no standalone binary

Common usage

# interactive mode
./dist/cli.js

# one-shot prompt
./dist/cli.js -p "what files are in this directory?"

# choose a model explicitly
./dist/cli.js --model claude-opus-4-6

# switch to Codex backend
CLAUDE_CODE_USE_OPENAI=1 ./dist/cli.js

Selected environment variables

Variable Purpose
ANTHROPIC_API_KEY Anthropic API key
ANTHROPIC_AUTH_TOKEN alternative Anthropic auth token
ANTHROPIC_MODEL override default Anthropic model
ANTHROPIC_BASE_URL custom Anthropic-compatible endpoint
CLAUDE_CODE_USE_OPENAI force OpenAI Codex backend
CLAUDE_CODE_USE_BEDROCK switch to AWS Bedrock
CLAUDE_CODE_USE_VERTEX switch to Google Vertex AI
CLAUDE_CODE_USE_FOUNDRY switch to Anthropic Foundry
CLAUDE_CODE_OAUTH_TOKEN OAuth token provided via environment
CLAUDE_CONFIG_DIR override the config directory (default: ~/.my-claude-code)

Config isolation

This fork defaults CLAUDE_CONFIG_DIR to ~/.my-claude-code so its settings, auth tokens, teammate models, and MCP configs are fully separate from an official Claude Code installation in ~/.claude.

To share config with the official installation:

export CLAUDE_CONFIG_DIR="$HOME/.claude"
./dist/cli.js

Feature Flags and Supporting Docs

  • FEATURES.md: technical audit of compile-time feature flags in this snapshot
  • AGENTS.md: Codex-oriented repo guidance for coding agents
  • CLAUDE.md: Claude-oriented repo guidance for coding agents
  • ARCHITECTURE.md: deep-dive into codebase design and architecture

The default build already includes this repo's current working feature bundle:

bun run build
./dist/cli.js

To enable specific flags manually:

bun run ./scripts/build.ts --feature=ULTRAPLAN --feature=ULTRATHINK

Project Structure

scripts/
  build.ts                build script and feature flag bundler

src/
  entrypoints/cli.tsx     CLI entrypoint
  commands.ts             slash command registry
  tools.ts                tool registry
  QueryEngine.ts          message and tool orchestration

  commands/               slash command implementations
  tools/                  tool implementations
  components/             Ink/React terminal UI
  hooks/                  React hooks
  services/               API, OAuth, MCP, analytics integrations
  state/                  application state
  skills/                 skill system
  plugins/                plugin system
  bridge/                 IDE bridge
  voice/                  voice support
  tasks/                  background task management

Contributing

Issues and pull requests are welcome. For larger changes, open an issue first to discuss the approach.

Yorumlar (0)

Sonuc bulunamadi