# How to Install Archify: A Complete Guide for AI Agent Integration

> Easily install Archify for AI agent integration using npx skills add. Integrate tt-a1i/archify globally and start using it with natural language prompts for Cursor, Claude Code, and Codex CLI.

- Repository: [tt-a1i/archify](https://github.com/tt-a1i/archify)
- Tags: getting-started
- Published: 2026-08-29

---

**Install Archify using `npx skills add tt-a1i/archify -g` for global access across Cursor, Claude Code, Codex CLI, and other supported agents, then invoke it with natural language prompts.**

Archify is a **Node.js skill** distributed as a self-contained npm module that enables AI agents to generate architecture diagrams from codebases. According to the `tt-a1i/archify` source code, the skill is designed to be **zero-dependency** aside from Node.js itself, making it safe for isolated environments. This guide walks you through the three-step installation process documented in the repository's README and skill definition files.

## Prerequisites

Before installing Archify, ensure your environment meets the baseline requirements. The skill requires only Node.js to operate—no additional runtime dependencies are bundled or required.

- **Node.js**: Required for the `npx` command execution and skill runtime
- **Supported AI Agent**: Cursor, Claude Code, Codex CLI, Opencode, or compatible MCP-based agents

## Installation Methods

The `tt-a1i/archify` repository supports multiple installation patterns depending on your use case, from permanent global installation to temporary execution.

### Global Installation (Recommended)

The standard approach registers Archify as a global skill accessible across all your projects. This method uses the `skills` CLI via `npx` to download the package and register it with your target agent.

```bash

# Global install for most agents (Cursor, Claude Code, etc.)

npx skills add tt-a1i/archify -g

```

This command pulls the skill package—containing the CLI entry point `archify/bin/archify.mjs`, JSON schemas, and HTML/SVG renderers—and registers it according to the formal definition in [`archify/SKILL.md`](https://github.com/tt-a1i/archify/blob/main/archify/SKILL.md).

### Explicit Non-Interactive Installation (CI/CD)

For automated environments or scripts where interactive prompts must be disabled, use the explicit flag set. This pattern is documented in the README for CI/CD pipelines requiring deterministic behavior.

```bash

# Non-interactive install for Cursor in automated environments

npx -y skills add tt-a1i/archify \
  --skill archify \
  --agent cursor \
  --global \
  --copy \
  --yes

```

The `--yes` flag auto-confirms prompts, while `--copy` ensures the skill files are copied locally rather than linked.

### Temporary Usage Without Installation

If you need to run Archify once without persisting the skill to your agent's configuration, use the `use` subcommand. This executes the skill immediately without registration.

```bash

# Run without installing (temporary use)

npx skills use tt-a1i/archify@archify --agent codex

```

## Configuring Your AI Agent

After installation, Archify requires agent-specific configuration to handle the generate→validate→render pipeline. The repository provides **agent-aware launch URLs** that pre-configure the correct global and project commands for each supported platform.

Supported agents include:
- **Cursor**
- **Claude Code**
- **Codex CLI**
- **Opencode**

The skill reads its configuration from [`archify/SKILL.md`](https://github.com/tt-a1i/archify/blob/main/archify/SKILL.md) to determine validation rules and renderer contracts stored in [`archify/references/viewer-runtime.md`](https://github.com/tt-a1i/archify/blob/main/archify/references/viewer-runtime.md).

## Running Archify

Once installed globally, invoke Archify through natural language prompts in your agent's chat interface:

```text
Use archify to map this repository's runtime architecture.

```

### Interactive Guide Mode

To explore available diagram types without connecting to an AI agent, run the CLI directly using Node.js:

```bash

# Launch the interactive guide

node archify/bin/archify.mjs guide "Show an API request with Redis cache miss"

```

This command executes the validation pipeline defined in the source: the CLI generates a typed JSON IR, runs validators (schema, layout, route, label checks), and compiles the output into a single HTML file.

## Understanding the Installation Architecture

The installation process sets up three core components that work together to ensure deterministic diagram generation:

**Skill Package**  
The npm module bundles `archify/bin/archify.mjs` (the CLI driver), JSON schema definitions in `archify/schemas/`, and rendering engines. As documented in [`archify/SKILL.md`](https://github.com/tt-a1i/archify/blob/main/archify/SKILL.md), this package is self-contained and requires no external runtime dependencies beyond Node.js.

**Validation Pipeline**  
When your agent generates a diagram specification, Archify executes a series of validators (schema, layout, route, and label checks) against the definitions in [`archify/schemas/README.md`](https://github.com/tt-a1i/archify/blob/main/archify/schemas/README.md). Only fully-validated artifacts proceed to rendering.

**Renderer**  
Validated JSON is compiled into a standalone HTML file with optional PNG, SVG, or WebM exports. The renderer contract is defined in [`archify/references/viewer-runtime.md`](https://github.com/tt-a1i/archify/blob/main/archify/references/viewer-runtime.md), ensuring consistent output across all supported agents.

## Summary

- **Install Archify** using `npx skills add tt-a1i/archify -g` for permanent access or `npx skills use` for temporary execution
- **Zero-dependency design** requires only Node.js, making it safe for isolated or containerized environments
- **Three-step workflow**: add the skill via `npx`, configure your specific AI agent (Cursor, Claude Code, etc.), and run via natural language prompts
- **Architecture** consists of the skill package (`archify/bin/archify.mjs`), validation pipeline (`archify/schemas/`), and HTML/SVG renderer
- **CI/CD support** available through explicit non-interactive flags (`--yes`, `--copy`, `--global`)

## Frequently Asked Questions

### What AI agents are compatible with Archify?

Archify supports Cursor, Claude Code, Codex CLI, Opencode, and other MCP-compatible agents. The repository provides agent-aware launch URLs in the quick-start documentation that pre-configure the correct command syntax for each platform.

### Does Archify require additional dependencies beyond Node.js?

No. According to the [`archify/SKILL.md`](https://github.com/tt-a1i/archify/blob/main/archify/SKILL.md) definition, the skill is **zero-dependency** aside from Node.js itself. The package bundles all necessary validators, renderers, and the CLI entry point (`archify/bin/archify.mjs`) in a single self-contained module.

### Can I install Archify in a CI/CD pipeline?

Yes. Use the explicit non-interactive installation method with flags: `npx -y skills add tt-a1i/archify --skill archify --agent cursor --global --copy --yes`. The `--yes` flag suppresses interactive prompts, and `--copy` ensures local file persistence rather than symbolic linking.

### Where is the skill configuration stored after installation?

Global installations register the skill with your agent's global configuration directory. The specific path varies by agent (Cursor vs. Claude Code), but the skill definition in [`archify/SKILL.md`](https://github.com/tt-a1i/archify/blob/main/archify/SKILL.md) and schema references in [`archify/schemas/README.md`](https://github.com/tt-a1i/archify/blob/main/archify/schemas/README.md) are copied locally during the `--copy` operation to ensure offline availability.