ChatGPT cannot send a static authorization header to an MCP server. That is a limit of the host, not of Agent Update, and no config works around it. Use the HTTP API below: the same five tools without the protocol.
Where it goes: Custom GPT -> Configure -> Actions (or an API function tool)
Setup
curl -sS -X POST https://api.tryagentupdate.com/v1/agent/messages \
-H "Authorization: Bearer au_live_..." \
-H "Content-Type: application/json" \
-d '{"text":"Build finished on main."}'Before you file a bug
ChatGPT does support remote MCP servers; the docs say SSE and streaming HTTP. But the auth dialog offers exactly three options: OAuth, No Authentication, and Mixed Authentication. There is no custom-header field and no API-key field, and OpenAI's docs are explicit that ChatGPT cannot present custom API keys. Do not invent a header workaround. There is none. Point a connector at /v1/mcp with "No Authentication" and it sends unauthenticated requests, so every tool call 401s. The only native alternative is adding full OAuth 2.1 to /v1/mcp — authorization code with PKCE, plus CIMD or dynamic client registration — which is a product decision, not a config snippet. Until that exists, publish the REST endpoint and call it from a Custom GPT Action or a function tool, where you control the headers.
Check it worked
Ask the agent to text you. A 401 means the token is wrong, or the word Bearer is in there twice. No tool at all means the transport string is wrong. Both are covered in troubleshooting.
Next
The five tools and their arguments are on the MCP page. The same tools over plain HTTP are on the REST page. Every other tool is on the docs index.