airc

AI Config Manager - A TypeScript-based CLI tool to manage and install AI tool configuration files across multiple AI development tools.

Idea is manage your AI tools config in same place and then reuse everywhere.

Installation

``bash
npm install airc-cli -g
airc --help

Or use npx directly


npx airc-cli --help
`

Quick Start

$3

Initialize your local config directory:

`bash
airc init
`

This creates:

- ~/.airc/ - User config directory
- ~/.airc/config/skills/ - Local skills config storage
- ~/.airc/config/commands/ - Local commands config storage
- ~/.airc/config/agents/ - Local commands config storage
- ~/.airc/config/mcp-config.json - MCP server configurations
- ~/.aircrc - Configuration file for airc cli

$3

Add a git repository containing AI configs. This automatically copies config files to your local directory:

`bash


Add from GitHub (full repository)

airc source add https://github.com/vercel-labs/agent-skills

Add with custom name

airc source add https://github.com/vercel-labs/agent-skills --name vercel-skills

Add from GitHub tree URL

airc source add https://github.com/vercel-labs/agent-skills/tree/main/skills/react-best-practices
`

What happens:

- Repository is cloned to ~/.airc/temp/
- Config folders (skills, commands, agents) are automatically discovered
- Config files are copied to ~/.airc/config/{type}/{source-name}/
- Metadata is tracked in ~/.aircrc for smart updates
- Remove temp repo

$3

List all configs in your local directory:

`bash


List all configs

airc list

Filter by type

airc list --type skills
`

$3

Install your configs to specific AI tools:

`bash


Install to Claude for your project only(/.claude)

cd
airc install claude

Install globally(~/.claude)

airc install claude --global
`

$3

`bash
airc source update react-best-practices
`

$3

`bash
airc remove react-best-practices
`

$3

Initially we have three folders in config folder(skills / commands / agents), but you can add your own custom folder, they will be copied to target folder when using airc install command:

`
airc/
├── config/ # Default configuration files
│ ├── commands/ # Command definitions
│ ├── skills/ # Skill definitions
│ ├── agents/ # Agent definitions
│ └── scripts/ # Add your own config folder
`

Managing MCPs (Model Context Protocol)

airc now supports managing MCP (Model Context Protocol) server configurations. MCPs allow AI tools to interact with external systems and services through standardized server interfaces.

$3

MCP configurations are stored in ~/.airc/config/mcp-config.json:

`json
{
"filesystem": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users"]
},
"github": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"],
"env": {
"GITHUB_PERSONAL_ACCESS_TOKEN": "your_token_here"
}
}
}
`

Field Descriptions:
- type: Protocol type (currently only "stdio" is supported)
- command: The command to run the MCP server (e.g., "npx", "node")
- args: Array of command arguments (optional)
- env: Environment variables for the server (optional)

$3

List all available MCPs along with your other configs:

`bash
airc list
`

Example output:
`
📋 Your Configs

📦 Commands (2)
✏️ manual commit
✏️ manual push

📦 MCP Configs (2)
✏️ manual filesystem
type: stdio
✏️ manual github
type: stdio

Total: 4 configs
0 imported, 4 manual
`

$3

When you run airc install , you'll be prompted to select which MCPs to install:

`bash


Install to Claude (project-level)

cd
airc install claude

Install globally

airc install claude --global
`

Interactive MCP selection:
`
📦 MCP Configs
? Select MCPs to install: (Press to select, to toggle all)
❯◯ filesystem (stdio)
◯ github (stdio)
`

After selection, MCPs are merged into the tool's config file:
`
✅ Successfully installed 2 MCP configs
Installed to: .mcp.json
`

$3

If an MCP with the same name already exists in the tool's config, use the --force flag to overwrite:

`bash
airc install claude --force
`

Without --force:
`
❌ Error: MCP conflicts detected: filesystem, github
These MCPs already exist in the config file.
Use --force to overwrite existing MCPs.
`

With --force:
`
✅ Successfully installed 2 MCP configs (overwrote 2 existing)
`

$3

Remove an MCP from your local configuration:

`bash
airc remove --force
`

Example:
`bash
airc remove github --force
`

Output:
`
🗑️ Remove Config

MCP Config: github
Type: stdio
Source: manual (MCP configuration)

✅ Successfully removed MCP "github"
MCP removed from ~/.airc/config/mcp-config.json
`

$3

Different AI tools store MCPs in different locations with different field names:

| Tool | Project Mode Path | Global Mode Path | Field Name |
|------|------------------|------------------|------------|
| Claude | .mcp.json | ~/.claude.json | mcpServers |
| Gemini | .gemini/settings.json | ~/.gemini/settings.json | mcpServers |
| Antigravity | N/A (global only) | ~/.gemini/antigravity/mcp_config.json | mcpServers |
| Cursor | .cursor/mcp.json | ~/.cursor/mcp.json | mcpServers |
| Copilot | .copilot/mcp-config.json | ~/.copilot/mcp-config.json | mcpServers |
| OpenCode | opencode.json | ~/.config/opencode/opencode.json | mcp |

Note: Antigravity only supports global mode. When installing to Antigravity in project mode, you'll see an informational message:

`
ℹ️ Note: Antigravity only supports global MCP installation
Use --global flag to install MCPs for Antigravity
`

$3

MCPs not showing in airc list:
- Ensure ~/.airc/config/mcp-config.json exists and contains valid JSON
- Check file permissions on the mcp-config.json file

MCP installation fails:
- Verify the tool supports MCP format (see table above)
- For Antigravity, use --global flag
- Check that the target config file has write permissions
- Use --force flag if MCPs already exist and you want to overwrite

MCPs not working after installation:
- Verify the MCP server command is correctly installed (e.g., npx @modelcontextprotocol/server-filesystem`)
- Check environment variables are set correctly in the MCP config
- Restart your AI tool after installing MCPs
- Consult the specific AI tool's documentation for MCP troubleshooting