ai-context-inspector

ai-context-inspector

Scans projects to discover all AI configuration files, MCP servers, agents, and memories across 19 tools, enabling export and import of the entire AI ecosystem.

Category
Visit Server

README

<p align="center"> <h1 align="center">@cocaxcode/ai-context-inspector</h1> <p align="center"> <strong>One command. Every AI config in your project. Instantly.</strong><br/> CLI + MCP server · 19 tools detected · Export/Import across 7 targets </p> </p>

<p align="center"> <a href="https://www.npmjs.com/package/@cocaxcode/ai-context-inspector"><img src="https://img.shields.io/npm/v/@cocaxcode/ai-context-inspector.svg?style=flat-square&color=cb3837" alt="npm version" /></a> <a href="https://www.npmjs.com/package/@cocaxcode/ai-context-inspector"><img src="https://img.shields.io/npm/dm/@cocaxcode/ai-context-inspector.svg?style=flat-square" alt="npm downloads" /></a> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/license-MIT-blue.svg?style=flat-square" alt="License" /></a> <img src="https://img.shields.io/badge/node-%3E%3D20-339933?style=flat-square&logo=node.js&logoColor=white" alt="Node" /> <img src="https://img.shields.io/badge/AI%20tools-19%20supported-blueviolet?style=flat-square" alt="19 AI tools" /> </p>

<p align="center"> <a href="#quick-start">Quick Start</a> · <a href="#mcp-server-setup">MCP Server</a> · <a href="#what-it-detects">What It Detects</a> · <a href="#features">Features</a> · <a href="#mcp-tool-reference">Tool Reference</a> · <a href="#architecture">Architecture</a> </p>


The most comprehensive AI ecosystem scanner available. AI Context Inspector discovers every AI configuration file, MCP server, custom skill, agent, and persistent memory in your project — across 19 tools — and gives you the full picture in seconds. Export your entire AI setup and import it into any of 7 supported tools (Claude, Cursor, Windsurf, Copilot, Gemini, Codex, OpenCode) with automatic secret redaction. Zero config. Zero runtime dependencies beyond MCP SDK and Zod. Everything runs locally — your configs and credentials never leave your machine.


Quick Start

npx -y @cocaxcode/ai-context-inspector@latest

That's it. An HTML dashboard opens with everything your AI tools know about your project.

# Scan a specific directory
npx -y @cocaxcode/ai-context-inspector@latest --dir /path/to/project

# Include user-level configs (~/.claude, ~/.gemini, etc.)
npx -y @cocaxcode/ai-context-inspector@latest --user

# Live-connect to each MCP server and list its tools
npx -y @cocaxcode/ai-context-inspector@latest --introspect

# Output raw JSON instead of HTML
npx -y @cocaxcode/ai-context-inspector@latest --json

<details> <summary>All CLI flags</summary>

Flag Description Default
--dir <path> Directory to scan . (current)
--user Include user-level configs (~/) false
--introspect Connect to MCP servers and list tools false
--timeout <ms> Introspection timeout per server 10000
--json Output JSON instead of HTML false
--mcp Run as MCP server (stdio transport) false

</details>


MCP Server Setup

Run as an MCP server so your AI tools can inspect their own ecosystem.

Claude Code

claude mcp add --scope user ai-context-inspector -- npx -y @cocaxcode/ai-context-inspector@latest --mcp

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "ai-context-inspector": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/ai-context-inspector@latest", "--mcp"]
    }
  }
}

<details> <summary>Cursor / Windsurf / VS Code / Codex CLI</summary>

Cursor / Windsurf — add to .cursor/mcp.json or equivalent:

{
  "mcpServers": {
    "ai-context-inspector": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/ai-context-inspector@latest", "--mcp"]
    }
  }
}

VS Code — add to .vscode/mcp.json:

{
  "servers": {
    "ai-context-inspector": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cocaxcode/ai-context-inspector@latest", "--mcp"]
    }
  }
}

Codex CLI:

codex mcp add ai-context-inspector -- npx -y @cocaxcode/ai-context-inspector@latest --mcp

</details>

Or add directly to any .mcp.json:

{
  "mcpServers": {
    "ai-context-inspector": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cocaxcode/ai-context-inspector@latest", "--mcp"]
    }
  }
}

What It Detects

Tool Context Files What's Found
Claude Code CLAUDE.md, .claude/, .mcp.json Instructions, MCP config, skills, agents, memory
Cursor .cursorrules, .cursor/rules/, .cursorignore Rules, ignore patterns
Windsurf .windsurfrules, .windsurf/rules/ Codeium/Windsurf rules
GitHub Copilot .github/copilot-instructions.md, .github/agents/ Instructions, custom agents
Gemini CLI GEMINI.md, .gemini/, .geminiignore Config, rules
OpenAI Codex AGENTS.md, AGENT.md, .codex/ Agent instructions
OpenCode OPENCODE.md, .opencode/, opencode.json CLI config
Aider .aider.conf.yml, .aiderignore Config, model settings
Cline .clinerules, .clineignore Rules (file or directory)
Roo Code .roo/rules/, .roorules, .rooignore Rules
Continue.dev .continuerules, .continue/config.yaml Rules, config
Amazon Q .amazonq/rules/ Developer rules
Augment .augment/rules/, .augment-guidelines Code rules
Replit .replit.md Agent instructions
Firebase Studio .idx/airules.md AI rules
VS Code .vscode/mcp.json MCP configuration
Tabnine .tabnine.yaml AI config
Sourcegraph .sourcegraph/ Cody config
Universal CONVENTIONS.md Multi-tool conventions

Tip: Pass --user to also scan user-level configs: ~/.claude/, ~/.gemini/, ~/.codex/, ~/.continue/, ~/.aider.conf.yml, ~/.augment/, ~/.github/agents/, ~/.codeium/, ~/.opencode/, ~/.tabnine/.

Beyond context files, the scanner also detects:

  • MCP servers from .mcp.json, ~/.claude.json, .vscode/mcp.json, Claude Desktop config — with optional live introspection
  • Custom skills in .claude/skills/ (including symlinks from npx skills add)
  • Custom agents in .claude/agents/ with memory detection
  • Persistent memories — Claude Memory, agent memory, engram, OpenSpec, ATL

Features

Scan

npx -y @cocaxcode/ai-context-inspector@latest
npx -y @cocaxcode/ai-context-inspector@latest --user --introspect

Generates a self-contained HTML dashboard with dark/light mode, search, collapsible sections, file previews, and color-coded tool badges. Pass --json for raw output.

Export

# Export your AI ecosystem to .aci/bundle.json
npx -y @cocaxcode/ai-context-inspector@latest export

# Export only MCP and context configs, redact secrets
npx -y @cocaxcode/ai-context-inspector@latest export --only mcp,context --secrets none

<details> <summary>Export flags</summary>

Flag Description Default
--dir <path> Directory to export .
--output <path> Output directory .aci/
--include-user Include user-level configs false
--only <categories> Filter: mcp,skills,agents,memories,context all
--secrets <mode> none (redact) or all (include) interactive

</details>

Import

# Import into a Cursor project
npx -y @cocaxcode/ai-context-inspector@latest import --target cursor

# Auto-detect tool, skip confirmation, overwrite existing
npx -y @cocaxcode/ai-context-inspector@latest import --yes --force

<details> <summary>Import flags</summary>

Flag Description Default
[file] Path to bundle JSON auto-detect .aci/bundle.json
--dir <path> Target directory .
--target <tool> Target tool (see table) auto-detect
--scope <scope> project or user per-resource
--force Overwrite existing resources false
--yes Skip confirmation prompt false
--only <categories> Filter: mcp,skills,agents,memories,context all
--secrets <mode> none or all interactive

</details>

Supported Import Targets

Target MCP Config Context File Rules Dir Skills Agents
claude .mcp.json CLAUDE.md -- .claude/skills/ .claude/agents/
cursor .cursor/mcp.json .cursorrules .cursor/rules/ -- --
windsurf .mcp.json .windsurfrules .windsurf/rules/ -- --
copilot .vscode/mcp.json .github/copilot-instructions.md .github/instructions/ -- .github/agents/
gemini .gemini/settings.json GEMINI.md .gemini/rules/ -- --
codex .mcp.json AGENTS.md .codex/rules/ -- --
opencode opencode.json OPENCODE.md .opencode/rules/ -- --

Note: Secrets handling has three modes: none (redact all — safe for sharing), all (include as-is), or interactive (prompt per variable, CLI default). Environment variables in MCP configs are detected automatically.


MCP Tool Reference

When running as an MCP server (--mcp), five tools are exposed:

scan

Discover the complete AI ecosystem in a project directory.

Parameter Type Required Description
directory string No Path to scan (default: cwd)
include_user boolean No Include user-level configs
introspect boolean No Live-connect to MCP servers
timeout number No Introspection timeout (ms)

introspect_mcp

Connect to a specific MCP server and list its tools, resources, and prompts.

Parameter Type Required Description
command string Yes* Command to launch the server
args string[] No Arguments for the command
env object No Environment variables
url string Yes* URL for HTTP/SSE servers
transport string No stdio (default) or http
timeout number No Connection timeout (ms)

*Either command (stdio) or url (http) is required.

generate_report

Generate a standalone HTML dashboard from scan results.

Parameter Type Required Description
directory string No Path to scan
include_user boolean No Include user-level configs
introspect boolean No Introspect MCP servers
timeout number No Introspection timeout (ms)

export_ecosystem

Export the complete AI ecosystem to a portable .aci/bundle.json.

Parameter Type Required Description
dir string No Directory to scan (default: cwd)
include_user boolean No Include user-level configs
only string[] No Categories: mcp, skills, agents, memories, context
secrets string No "none" (default), "all", or ["VAR1", "VAR2"]

import_ecosystem

Import a bundle into a project, adapting configuration to the target tool.

Parameter Type Required Description
file string No Path to bundle (auto-detects .aci/bundle.json)
dir string No Target directory (default: cwd)
target string No Target tool (auto-detects): claude, cursor, windsurf, copilot, gemini, codex, opencode
scope string No project or user
force boolean No Overwrite existing resources
confirm boolean No Execute import (default: false = dry-run)
only string[] No Categories to import
secrets string No "none", "all", ["VAR1"], or {"VAR1": "value"}

Warning: import_ecosystem defaults to dry-run mode (confirm: false). Set confirm: true to actually write files.


Architecture

src/
├── index.ts              # Entry: CLI vs MCP mode routing
├── cli.ts                # CLI arg parsing + orchestration
├── server.ts             # createServer() MCP factory
├── scanner/              # 6 parallel scanners (Promise.all)
│   ├── catalog.ts        # AI_FILE_CATALOG — 50+ entries, 19 tools
│   ├── context-files.ts  # Scan .md, rules, configs
│   ├── mcp-configs.ts    # Parse .mcp.json, ~/.claude.json, etc.
│   ├── mcp-introspect.ts # Connect to servers, list tools/resources
│   ├── skills.ts         # Skills + symlinks + frontmatter
│   ├── agents.ts         # Agents + memory detection
│   └── memories.ts       # engram, openspec, .atl, claude memory
├── ecosystem/            # Export/import engine
│   ├── export.ts         # Scan → bundle → .aci/bundle.json
│   ├── import.ts         # Load → plan → execute
│   ├── target-map.ts     # Path configs for 7 AI tools
│   ├── detect-target.ts  # Auto-detect tool in project
│   └── secrets.ts        # Env var detection + redaction
├── report/               # HTML dashboard generator
│   ├── generator.ts      # generateHtml(ScanResult) → string
│   ├── sections.ts       # Section renderers
│   ├── styles.ts         # CSS (dark/light)
│   └── scripts.ts        # JS (collapse, search, preview)
└── tools/                # 5 MCP tool handlers
    ├── scan.ts
    ├── introspect.ts
    ├── report.ts
    ├── export.ts
    └── import.ts

Stack: TypeScript 5 (strict, ESM) / MCP SDK / Zod / tsup / Vitest


MIT · Built by cocaxcode

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured