MCP server exposing OpenAI gpt-image-2 image generation and editing as a single tool for Claude Code and other MCP clients.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"mcp-openai-images-audio","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# mcp-openai-images-audio [`v0.1.0`](https://github.com/eduard256/mcp-openai-images-audio/releases/tag/v0.1.0) MCP server that exposes OpenAI's `gpt-image-2` and `gpt-image-1.5` to Claude Code as a single tool. Generate, edit, or compose images straight from a chat. Files are written to disk; the model never returns base64 to your context. This isn't a wrapper for everything OpenAI does. It's one tool: `image`. That's intentional.  One call, `size: 2048x1152`, `quality: high`, no references — produced this UI mockup. Real readable typography, real-looking code preview, accurate Instagram visual language. That's the level you should expect. ## Install ```bash pip install mcp-openai-images-audio ``` Or run directly without installing: ```bash uvx mcp-openai-images-audio ``` ## Connect to Claude Code ```bash claude mcp add openai-images \ --scope user \ -e OPENAI_API_KEY=sk-... \ -- uvx mcp-openai-images-audio ``` **Important:** 1. **Organization verification is mandatory** for `gpt-image-2`. Verify at https://platform.openai.com/settings/organization/general. Takes a few minutes, propagates within 15 minutes. 2. The API key needs billing credit. Without it you get a 400 with `billing_hard_limit_reached`. 3. The server runs over stdio. No HTTP, no separate process to keep alive — Claude Code starts and stops it for you. ## How the tool works One tool, three modes selected by `references_paths`: - empty / not passed → `/v1/images/generations` (text → new image) - 1 path → `/v1/images/edits` (modify that image) - 2..16 paths → `/v1/images/edits` (compose with labeled references) The server picks the model on its own: - `background='transparent'` → `gpt-image-1.5` (gpt-image-2 currently rejects alpha — confirmed regression in OpenAI's docs) - everything else → `gpt-image-2` The ac…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.