MCP servers · ai
OpenAI MCP server
AI completions, embeddings, image generation, TTS. Hosted by 0n3: 4 endpoints your AI can call with your own OpenAI account, from Claude, ChatGPT, Cursor, Claude Code, Codex or Gemini CLI. One sign-in, one key per AI, a receipt for every call.
4 endpoints5 categoriesauth: api_keyhosted
What your AI can do with OpenAI
Generate image · Generate images with DALL-E1
Create an image
POST /images/generationsCreate embedding · Create text embeddings1
Create an embedding
POST /embeddingsText to speech · Convert text to speech1
Text to speech
POST /audio/speechGenerate text · Generate text completions0
Generate text completions
More · Everything else this app exposes.1
Chat completion
POST /chat/completionsInstall in three steps
- Create a free account at 0n3.app and connect OpenAI: paste a key or sign in through a door. It lands in your vault, encrypted.
- Open 0n3.app/install, pick your AI. A key is made for that AI alone; copy the line it gives you.
- Ask your AI: whoami. It answers with your connected services. Then ask it to create an image in OpenAI.
Install (user scope — every project)
claude mcp add --transport http --scope user 0n3 https://0n3.app/mcp --header "Authorization: Bearer <your 0n token>"
Connect OpenAI to
Claude Code + OpenAIMCP server + skillClaude (desktop & claude.ai) + OpenAIconnectorChatGPT + OpenAIconnector · GPT actionCodex CLI + OpenAIMCP serverGemini CLI + OpenAIMCP serverCursor + OpenAIMCP serverVS Code (Copilot) + OpenAIMCP serverWindsurf + OpenAIMCP server
More ai MCP servers
Questions
What can an AI do with the OpenAI MCP server?
4 things: generate image, create embedding, text to speech, generate text and more. Every call runs with your own OpenAI connection and lands on a receipt.
Do I need to run anything on my computer?
No. 0n3 is hosted. Connect OpenAI once, install 0n3 into your AI with one line, done.
Where does my OpenAI key go?
Into your 0n3 vault, encrypted, one record per connection. The AI never sees it.