ZCF - Zero-Config Code Flow
Zero-config, one-click setup for Claude Code & Codex with bilingual support, intelligent agent system, and personalized AI assistant.
Evidence: The CLI tool performs operations via interactive menus and command-line arguments, requiring user input for API keys (e.g., `-k` parameter), but does not explicitly state least-privilege principles. User confirmation exists (interactive menus), but non-interactive mode may execute automatically. Data flow transparency is insufficient; no detailed explanation of how data is processed or transmitted. Sensitive data handling: API keys are passed as arguments, but storage or protection measures are not described. Dependency security: dependencies are managed via pnpm catalog, but no dependency audit or vulnerability scanning evidence is provided. External effects: the tool modifies user configuration files (e.g., installing workflows), but does not clearly prompt or obtain consent. Rollback: no explicit rollback mechanism is provided. Source attribution: README acknowledges multiple open-source projects, but detailed code provenance is not provided. Deductions: least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, and rollback lack detailed evidence or implementation descriptions.
Evidence: Code structure is clear, tests cover edge cases and error handling, but tests are mostly mocked and do not verify real environment behavior. Dependency availability: uses `npx` to call external tools (e.g., ccusage), relying on network and external service availability. Failure messages: error handling functions (e.g., handleGeneralError) provide error output, but specific message content is not shown in provided files. Deductions: reliance on external services without offline or fallback options; specific failure message content not verified.
Evidence: README provides multiple usage scenarios (e.g., `npx zcf i`, `npx zcf u`), supports multiple languages (Chinese, English, Japanese), and provides non-interactive examples. Capability boundaries: does not clearly state the scope of supported or limited features. Trigger precision: command arguments (e.g., `-s`, `-p`) are documented, but not all options are detailed. Environment fit: supports multiple platforms (CI tests cover Windows, macOS, Linux), but specific environment requirements are not stated. Deductions: capability boundaries and trigger precision descriptions are incomplete.
Evidence: README provides quick start, full documentation links, sponsor list, community links, etc., with clear information architecture. Install notes are simple (`npx zcf`), but detailed installation steps or system requirements are not provided. Naming stability: project name and commands are stable, but version number is not explicitly stated in README. Examples and FAQ: basic command examples are provided, but FAQ is missing. Known limitations: not explicitly listed. License is MIT, versioning and changelog exist (CHANGELOG.md), maintenance responsibility is clear (author and contributors). Deductions: known limitations not stated, install notes not detailed enough.
Evidence: The tool provides one-click configuration, simplifying the setup process for Claude Code and Codex, with output as configuration files and workflows, which is practical. Marginal value: compared to manual configuration, it saves time but relies on external services. Cost-benefit: free and open-source, but may incur API call costs. Deductions: no data or user feedback on actual usage effectiveness.
Evidence: Claims in README (e.g., features, sponsors) have links, but detailed implementation documentation is not provided. Cross-source verification: tests and CI configuration provide some verification, but not independent. Fact-inference separation: README distinguishes facts (e.g., commands) from inferences (e.g., 'intelligent'), but not clearly. Deductions: claims lack detailed evidence, cross-source verification insufficient.
- API keys are passed via command-line arguments, which may be exposed in shell history or process lists; consider using environment variables or config files.
- The tool modifies user configuration files; in non-interactive mode, ensure clear prompts and user consent.
- Relies on external services (e.g., npx ccusage), which may fail without network; consider offline or fallback options.
What does this agent do, and when should you use it?
ZCF (Zero-Config Code Flow) is a CLI tool for Claude Code and Codex that simplifies the setup of AI coding environments by automating initialization, workflow configuration, API/CCR integration, and MCP settings. It is primarily used via npx zcf commands, offering an interactive menu and preset configurations, with support for Chinese, English, and Japanese interfaces. The project draws inspiration from community projects like CCR and the BMad Method, and integrates several API relay services as sponsors. ZCF is aimed at developers who want to quickly configure and optimize their AI coding workflows, reducing manual configuration overhead.
ZCF provides different operations through its CLI: npx zcf i performs a full initialization, including installation, workflows, API/CCR, and MCP (Model Context Protocol) setups; npx zcf u updates only the workflows; npx zcf --lang zh-CN switches the interface language. It also supports non-interactive mode, such as npx zcf i -s -p 302ai -k "sk-xxx", to specify providers and API keys. The tool reads the current project and generates or modifies configuration files to integrate with Claude Code and Codex. ZCF may call external API services and supports multiple sponsored API relays.
- A developer setting up Claude Code for the first time wants a quick environment setup without manually configuring API keys and workflows each time.
- A team wants to standardize Claude Code configurations across members to ensure consistent workflows.
- A user wants to switch between different API providers (e.g., 302.ai, AICodeMirror) to optimize cost or performance.
- A Codex user wants to simplify the initialization process and automatically configure associated MCP servers.
- A Chinese or Japanese developer prefers to use the tool in their native language.
- A CI/CD pipeline needs to automate the deployment of an AI coding environment using non-interactive scripts.
What are this agent's strengths and limitations?
- Zero-config one-click initialization saves significant setup time.
- Supports multiple API providers with presets and custom integration.
- Comprehensive documentation and an interactive CLI make it beginner-friendly.
- Multilingual support (Chinese, English, Japanese) appeals to international developers.
- Integrates community best practices (e.g., CCR, BMad Method) to enhance workflows.
- Specifically targets Claude Code and Codex, with limited support for other tools.
- Depends on external API providers; some features may require paid services.
- Initialization may modify project files, requiring careful handling.
- Documentation and community support are primarily via GitHub and Telegram.
How do you install or deploy this agent?
You can install via npm or use npx directly. The recommended way is to run npx zcf to open the interactive menu without global installation. Alternatively, install globally with: npm install -g zcf.
How do you use this agent?
Run npx zcf to open the interactive menu and select what you need. Common commands: npx zcf i for full initialization, npx zcf u to update workflows only. To switch language, use npx zcf --lang zh-CN. For non-interactive usage: npx zcf i -s -p 302ai -k "sk-xxx". Requires Node.js and network connectivity.
How does this agent compare with similar options?
Unlike manual configuration in Claude Code, ZCF automates the entire setup. It is similar to CCR (Claude Code Router) but focuses on zero-config and an all-in-one approach.