Before you begin
Have EvoLink and ChatGPT accounts, plus a client version, plan, and workspace permission that allows custom MCP. If the add action is unavailable, check host access first rather than running terminal instructions in chat.Connect EvoLink step by step
1
Open the custom MCP action
Open ChatGPT Plugins, choose Add or the plus button, then Create/Add custom MCP server. This is the website’s setup entry. Labels can vary by version; check workspace access if it is missing.
2
Enter the connection details
Use remote HTTP and keep the trailing
/mcp. Review the connection prompt and select Create as a plugin.3
Install and authorize the account
Open EvoLink in your personal plugins and install/connect it. When authentication is requested, sign in to EvoLink in the browser, review the account and permissions, and approve. Keep the flow open, then return to ChatGPT. Creating the configuration and installing the plugin are separate steps; the authorization action may appear at different stages.
4
Enable it in a conversation and verify
Start a new conversation and select EvoLink with
@. If your interface has Chat / Work tabs, switch to Work before starting the plugin conversation. Send the free verification prompt below.Success: actual account balance and available models, not only a statement that setup is complete.Free verification and the first media task
Once connected, send this in a conversation with EvoLink enabled:check_balance and search_models calls. Sign-in and these queries do not create paid media tasks. For generation, describe content and specifications, review the proposed model, parameters, and estimate, then explicitly approve submission. See the first media workflow.
Use publicly accessible reference links supported by the model. A chat attachment may not expose original file bytes to MCP, and the remote service cannot read your computer’s paths. See references and uploads.