Connect ChatGPT to Son

Son by Scorton is an agentic audit & advisory orchestrator. Agents execute. Humans exercise judgment. Son orchestrates. Scorton enforces trust.

Note to ChatGPT: you cannot add a connector yourself. Walk the user through the steps below one at a time with these exact values. Once connected, call son_whoami. A plain-text version is at /chatgpt.md.

Values

Connector name: Son by Scorton

MCP server URL: https://agents.scorton.dev/mcp

Authentication: OAuth — one-time code sent to your email

Steps

In ChatGPT, open Settings → Apps & Connectors → Advanced settings and turn on Developer mode. (Needs a Plus, Pro, Business, Enterprise or Edu plan; on a workspace plan an admin may have to allow custom connectors. Menu names can differ slightly by plan.)
Back in Apps & Connectors, choose Create (sometimes labelled “Add custom connector” or “New app”).
Fill in — Name: Son by Scorton · MCP server URL: https://agents.scorton.dev/mcp · Authentication: OAuth. Confirm you trust this server, then Create.
Click Connect. A Son window opens: enter your email, type the 6-digit code Son emails you, check that the screen says it will link ChatGPT, then press “Link ChatGPT to Son”.
Start a new chat, enable Son from the + / tools menu, and say: “Son, who am I?”

What Son can do

son_whoami — show your Son account and linked channels
son_list_specialists — list Son’s specialist sub-agents and their roles
son_start_vendor_review — run the flagship Vendor Security Review (synthetic Aperture LLM Gateway corpus); returns an AI Recommendation and a link for your human decision
son_get_result / son_list_work_items — follow up on reviews
son_request_human_decision — get the secure decision link for a pending review
son_mail_status — check the hello@scorton.cloud mail channel
search / fetch — read-only lookup of your WorkItems

Trust rules

Son is one orchestrator with specialist sub-agents (research, risk, security, compliance, evidence, reviewer, QA).
Son’s output is an AI Recommendation. It is never a Human Decision.
Final decisions are recorded only by you, on a Son web page, after confirming a fresh code sent to your email. The assistant cannot decide for you.
L5 (fully autonomous) operation is prohibited. Sub-agents cannot approve their own consequential actions.
Linking is explicit: Son links this assistant to your Son account only after you verify your email and approve the link screen.

Share with a teammate

Hey ChatGPT, connect me with Son. Open this exact link and follow its instructions: https://agents.scorton.dev/chatgpt

Son also answers by email at hello@scorton.cloud.