Install ImageKit Agent Skills and MCP configuration for your AI coding tools — one command, zero config.
Works with GitHub Copilot, Claude Code, Cursor, Codex, and any Agent Skills compatible client.
Run this in your project directory:
curl -fsSL https://raw.githubusercontent.com/shivamik/example-skill/main/install.sh | bashThat's it. You now have ImageKit skills available in your AI coding tool.
| Skill | What it does |
|---|---|
search-docs |
Search ImageKit documentation — ask about features, APIs, SDKs, transformations |
transformation-builder |
Build image/video transformation URLs from natural language |
By default, skills are installed to .agents/skills/ in your project — the universal Agent Skills standard directory recognized by all compatible clients.
curl -fsSL https://raw.githubusercontent.com/shivamik/example-skill/main/install.sh | bashFor global install (all projects):
curl -fsSL https://raw.githubusercontent.com/shivamik/example-skill/main/install.sh | bash -s -- --globaluvx --from "git+https://github.com/shivamik/example-skill.git" imagekit-skills installOr install persistently:
uv pip install "imagekit-skills @ git+https://github.com/shivamik/example-skill.git"
imagekit-skills installpip install "imagekit-skills @ git+https://github.com/shivamik/example-skill.git"
imagekit-skills install# Install skills to current project (.agents/skills/)
imagekit-skills install
# Install globally — available across all projects
imagekit-skills install --globalIf you want client-specific directories and MCP config:
imagekit-skills install --client copilot # .github/skills/ + .vscode/mcp.json
imagekit-skills install --client claude # .claude/skills/ + .claude/settings.json
imagekit-skills install --client cursor # .cursor/skills/ + .cursor/mcp.json
imagekit-skills install --client codex # .agents/skills/ + .codex/config.json
imagekit-skills install --client all # all of the aboveWhen you use
--client, the CLI also configures the ImageKit MCP server in that client's config so your AI tool can call ImageKit APIs directly.
imagekit-skills install search-docs
imagekit-skills install transformation-builderimagekit-skills install --dry-runimagekit-skills list # show available skills
imagekit-skills status # show what's installed where
imagekit-skills status --client claude # check a specific client
imagekit-skills remove # remove all skills
imagekit-skills remove search-docs # remove a specific skill
imagekit-skills --help # full help| Flag | Description |
|---|---|
--global |
Install to your home directory (~/) instead of the current project |
--client <name> |
Target a specific client: copilot, claude, cursor, codex, or all |
--skills-only |
Install skills but skip MCP config |
--mcp-only |
Configure MCP but skip skills |
--dry-run |
Show what would happen without making changes |
imagekit-skills install
│
▼
┌──────────────────────────┐
│ Copies SKILL.md files to │
│ .agents/skills/ │
└──────────────────────────┘
│
▼
Your AI tool discovers the skills automatically
│
▼
You type: "search ImageKit docs for overlay examples"
│
▼
AI uses the search-docs skill → calls ImageKit MCP → returns answer
When you use --client, it also writes MCP server config so your AI tool knows how to reach the ImageKit API:
{
"imagekit_api": {
"url": "https://api-mcp.imagekit.in/sse"
}
}After installing, just ask your AI tool naturally. The search-docs skill activates and searches ImageKit documentation for you.
The transformation-builder skill activates and uses the ImageKit MCP to build the transformation URL.
| Client | Skills Directory | MCP Config |
|---|---|---|
| GitHub Copilot (VS Code) | .github/skills/ |
.vscode/mcp.json |
| Claude Code | .claude/skills/ |
.claude/settings.json |
| Cursor | .cursor/skills/ |
.cursor/mcp.json |
| OpenAI Codex | .agents/skills/ |
.codex/config.json |
| Any Agent Skills client | .agents/skills/ |
— |
Remove skills from current project:
imagekit-skills removeRemove globally:
imagekit-skills remove --globalUninstall the CLI itself:
pip uninstall imagekit-skills