Car photos right from a chat.
Connect ChatGPT, Claude Code or another AI agent to your organization’s account. The agent takes photos from links, shows free previews and brings back the finished frames. Frames are paid from the organization’s balance at the same prices as in the dashboard.
How to connect
Pick your app. The server address is the same for all of them.
https://auto.happyin.app/mcp- ChatGPT
Through a custom MCP server in “Plugins”. Availability follows the policies of your account and workspace.
Steps
- Open chatgpt.com/plugins, press “+” and choose “Add custom MCP server”.
- Enter a name and a description, for example “Happyin Auto”.
- Under “Connection”, paste the server address (it ends in /mcp).
- In the authentication settings choose OAuth sign-in, confirm the warning and press “Create as a plugin”.
- Install the plugin. The auto.happyin.app sign-in page opens: sign in as the organization’s owner and allow access.
- In a new chat, type “@” and pick the plugin.
- Claudeclaude.ai and the Claude appsSoon
We are switching on sign-in for Claude in the browser and in the apps. Until then, connect Claude Code or any client with a key.
- Claude Codeterminal
One command, then sign in in the browser.
Steps
- Add the server with the command below.
- Start Claude Code, type /mcp, pick “happyin-auto” and sign in in the browser as the organization’s owner.
terminalclaude mcp add --transport http happyin-auto https://auto.happyin.app/mcpOr with a key
With a key instead of signing in: put the key in the environment variable AUTO_API_KEY (do not name it ANTHROPIC_…) and add the server with a header.
terminal (bash, zsh, PowerShell)claude mcp add --transport http happyin-auto https://auto.happyin.app/mcp --header 'Authorization: Bearer ${AUTO_API_KEY}' - Cursor and VS Codeeditors
Connect with a key: the Authorization header goes into mcp.json. Keep the key itself in an environment variable or enter it when asked.
Steps
- Cursor: create ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project) with the block below and put the key in the variable AUTO_API_KEY. Switch the server on in Customize.
- VS Code: run “MCP: Add Server” and choose HTTP, or add the block below to your user or project mcp.json (.vscode/mcp.json). The editor asks for the key on first start.
- Ask the agent for the organization’s balance: that shows the server works.
~/.cursor/mcp.json{ "mcpServers": { "happyin-auto": { "url": "https://auto.happyin.app/mcp", "headers": { "Authorization": "Bearer ${env:AUTO_API_KEY}" } } } }.vscode/mcp.json{ "servers": { "happyin-auto": { "type": "http", "url": "https://auto.happyin.app/mcp", "headers": { "Authorization": "Bearer ${input:auto-key}" } } }, "inputs": [ { "type": "promptString", "id": "auto-key", "description": "Happyin Auto API key", "password": true } ] } - Any MCP client, with a keyCodex, Gemini CLI and others
Any client that can send a header will do: the server address and an Authorization header with an API key.
Steps
- Get an API key. A test key (auto_sandbox_…) is free, and its results carry a watermark.
- In the client, set the server address and the header “Authorization: Bearer <key>”.
- Codex CLI reads the key from an environment variable; in Gemini CLI the header is part of the command, and the key is saved in its settings file.
headerAuthorization: Bearer auto_live_…~/.codex/config.toml[mcp_servers.happyin-auto] url = "https://auto.happyin.app/mcp" bearer_token_env_var = "AUTO_API_KEY"terminal (bash, zsh)gemini mcp add --transport http --header "Authorization: Bearer $AUTO_API_KEY" happyin-auto https://auto.happyin.app/mcpPowerShellgemini mcp add --transport http --header "Authorization: Bearer $env:AUTO_API_KEY" happyin-auto https://auto.happyin.app/mcp - API and webhooks
Your own scripts and systems: one request per frame, batches of up to 10,000, webhooks when they are ready.
To the API page - Install on your own servers
Photos stay in your network. Licence, rollout and support.
About the install
- Who connects
- The organization’s owner. They sign in to auto.happyin.app and allow the agent to work for the organization. The organization needs the “API” way of working switched on.
- Money
- The agent spends the organization’s balance at the same prices as the dashboard. Before anything that costs money it has to state the price and confirm the charge. Keep your app’s permission prompts for tool calls on.
- Photos by link
- Give the agent links to the photos (cloud storage, CDN, a site). A picture pasted into the chat reaches the model, not the tool.
- How to disconnect
- Dashboard → Organization → “Connected agents” → “Disconnect”: access ends at once. Changing your password disconnects the agents you connected.
What the agent can do
- Check the balance, backgrounds, processing modes and brand rules.
- Process one photo from a link and return the link to the result.
- Build a batch from links, show the free previews and send it to processing.
- Follow a batch, retry the frames that failed, or stop the batch.
It cannot delete batches or change brand rules and webhooks.