LLM Integration Guide

Make AI assistants aware of Inliner.ai, then give them live generation, editing, and project tools.

Machine-Readable Files

Following the llms.txt standard, we provide these files for AI systems:

Agent-Aware Installation

Inliner uses two complementary layers. The inliner-ai skill/plugin teaches the model when to activate Inliner and which workflow is safe. @inliner/mcp-server provides the 12 live tools and publishes the same tool-selection guidance during initialization.

OpenAI Codex — skill and MCP together

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

Gemini CLI extension

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

Claude Code and other skill-compatible agents can install skills/inliner-ai from the canonical agent-skill repository. Cursor users can install the dedicated Inliner Cursor plugin.

URL Format

The core concept: use stable project image URLs in your code, with authenticated tools creating first renders for account-owned namespaces.

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

Example:
https://img.inliner.ai/demo/happy-dog-playing-fetch_800x600.png

Integration Methods

• Direct URL — Use generated CDN URLs directly in HTML/CSS
• Agent Skill / Plugin — Teach Codex, Claude, Gemini, Cursor, and compatible agents when and how to use Inliner
• MCP Server — For Codex, Claude Code, Cursor, Copilot, Gemini CLI, and Windsurf
• CLI — npx @inliner/cli "sunset-beach-palm-trees_1200x600.png"
• REST API — Programmatic access via api.inliner.ai

Edit vs Generate Modality

For image changes, route to edit workflows with source context. Requests like "make it 600x600" can be interpreted as new generation if no source URL/path is provided.

• Edit existing image: Use edit_image (MCP) or inliner-ai edit ... (CLI) with explicit source
• Generate new image: Use generate_image (MCP) or inliner-ai generate ... (CLI), and wait for the completed asset before insertion
• Recommend URL only: Use recommend_image_url; it does not generate an asset
• If ambiguous: Ask whether to edit the previous image or create a new one

Cursor MCP Config (Recommended)

{
  "mcpServers": {
    "inliner": {
      "command": "npx",
      "args": ["-y", "@inliner/mcp-server"],
      "env": {
        "INLINER_API_KEY": "${INLINER_API_KEY}",
        "INLINER_DEFAULT_PROJECT": "your-project-namespace"
      }
    }
  }
}

Project-Specific Instructions

Users can get customized LLM instructions for their specific projects:

1. Log into the Inliner dashboard
2. Go to Projects
3. Click LLM Instructions on any project
4. Copy instructions optimized for Claude, Cursor, ChatGPT, etc.

These instructions include the user's namespace, custom prompts, and reference images.

Detailed Documentation

• MCP Server Setup Guide
• CLI Reference
• REST API Documentation
• Claude AI Integration
• Cursor AI Integration