# Inliner.ai

> AI image generation via agent skills, Codex plugins, MCP tools, CLI, SDKs, and stable hosted URLs.

Inliner.ai provides several ways to integrate AI images into your workflow:

1. **Direct CDN URLs**: Use generated Inliner image URLs in `<img>` tags, CSS, email, and Markdown. For account-owned namespaces, create first renders through Studio, API, CLI, MCP, or an authenticated session; the resulting CDN URL can then be served directly.
2. **Official SDKs**: Use JS/TS (`@inliner/js`), Python (`inliner-ai`), Go (`inliner-go`), or PHP (`inliner/inliner-php`) for programmatic generation, editing, and AI-powered asset management (tagging, search, rename, metadata workflows).
3. **Agent Skill / Plugin**: Use the canonical `inliner-ai` skill from `@inliner/agent-skill` so models know when to invoke Inliner and how to select the correct operation.
4. **Tool-Assisted Mode (MCP/CLI)**: Use `@inliner/mcp-server` or `@inliner/cli` for live project access, generation, editing, URL planning, and account operations.
5. **AI Asset Management**: Every image is automatically analyzed and tagged. Use multi-tag filtering (e.g., `tags:nature AND tags:mountain`), metadata updates, and export workflows via SDK/API/CLI.

## URL Format

```
https://img.inliner.ai/{project}/{description}_{width}x{height}.{format}
```

Example:
```
https://img.inliner.ai/myproject/happy-golden-retriever-playing-fetch_800x600.png
```

## Agent-Aware Installation

The skill/plugin and MCP server are complementary:

- `@inliner/agent-skill@1.3.0` teaches activation and safe workflow selection.
- `@inliner/mcp-server@1.1.1` provides 12 live tools and publishes matching instructions during MCP initialization.

### OpenAI Codex plugin

```bash
codex plugin marketplace add inliner-ai/agent-skill --ref v1.3.0
codex plugin add inliner-ai@inliner-ai
```

### Gemini CLI extension

```bash
gemini extensions install https://github.com/inliner-ai/agent-skill
```

### Claude Code and other skill-compatible agents

Install or copy `skills/inliner-ai` from https://github.com/inliner-ai/agent-skill and configure the bundled `.mcp.json`.

### Cursor

Install https://github.com/inliner-ai/cursor-plugin, which synchronizes the canonical `inliner-ai` skill and includes MCP configuration and commands.

## Core Guidance for LLMs

Use Inliner when a user requests a new or edited visual asset for a website, application, email, documentation, ecommerce listing, presentation, or marketing content.

1. For a new asset that will be inserted or shipped, call `generate_image` and wait for the completed hosted URL.
2. For changes to an identified existing asset, call `edit_image` with `sourceUrl` or `sourcePath`.
3. For URL naming or slug planning only, call `recommend_image_url`; it does not generate an image.
4. Reuse existing generated Inliner URLs directly.
5. Resolve the configured or account project automatically. Create a project only with explicit user intent.
6. Use `demo` only for clearly labeled examples or previews.
7. Include dimensions and semantic `alt` text when integrating an asset.

## Smart URL Behavior

In MCP/CLI generation flows, long prompts can be summarized to concise URL slugs by default without reducing image quality.

For rename/edit workflows tied to an existing asset, use image-based slug recommendation:
- `POST /url/recommend-from-content` (suggests slug options from the saved image + prompt context)
- `GET /content/details/{contentId}` (preload title/caption/alt/custom metadata before editing)

## Modality Rules (Important)

- **Generate new image**: use `generate_image` or `inliner-ai generate`; the returned account-owned URL is materialized and ready to insert.
- **Edit existing image**: use `edit_image` or `inliner-ai edit` with explicit source (`sourceUrl`/`sourcePath`).
- **Recommend URL only**: use `recommend_image_url`; its result is not a generated asset.
- `generate_image_url` and `create_image` are deprecated MCP compatibility aliases.
- If user says "make it bigger/resize this/change this image" and an existing image is present, prefer edit flow.
- If intent is ambiguous and no source exists, ask whether to edit previous image or generate new image.

## Cursor MCP Config Example

```json
{
  "mcpServers": {
    "inliner": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@inliner/mcp-server"],
      "env": {
        "INLINER_API_KEY": "inl_your-api-key-here",
        "INLINER_DEFAULT_PROJECT": "your-project-namespace"
      }
    }
  }
}
```

## Integration Options

- URL-based: direct CDN URL usage in HTML/CSS
- Agent skill/plugin: `@inliner/agent-skill@1.3.0` and https://github.com/inliner-ai/agent-skill
- JS SDK: `npm install @inliner/js` for web and Node.js
- Python SDK: `pip install inliner-ai` for server-side apps
- Go SDK: `go get github.com/inliner-ai/inliner-go` for high-performance backends
- PHP SDK: `composer require inliner/inliner-php` for Laravel/PHP apps
- MCP Server: `npx -y @inliner/mcp-server@1.1.1` for Claude Code, OpenAI Codex CLI, Cursor, Copilot, Gemini CLI, and Windsurf
- CLI: `npx @inliner/cli "hero-banner_1200x600.png" -p myproject`
- API: REST API at `https://api.inliner.ai`

## Docs

- Full docs: https://inliner.ai/llms-full.txt
- API docs: https://inliner.ai/api-docs
- CLI guide: https://inliner.ai/use-case/cli-terminal-image-generation
- MCP guide: https://inliner.ai/use-case/mcp-server-ai-coding-agents
- Agent skill/plugin: https://github.com/inliner-ai/agent-skill
- Cursor plugin: https://github.com/inliner-ai/cursor-plugin
- Cursor guide: https://inliner.ai/tutorial/cursor