What is an AI Keyholder?
An AI keyholder is a tool that reads the state of a chastity session and — strictly within predefined rules — issues directives. It fits two situations that stand side by side as equals. If the sub has a human keyholder, the AI takes the routine work off their hands and acts by their rules. If the sub has no keyholder, they can connect the AI as their keyholder themselves: it then takes on the keyholder function under rules the sub sets for themselves. This gives even someone without a partner reliable control — instead of making every lock period, every inspection and every decision by hand.
One thing first: responsibility stays with the human. The AI keyholder acts solely within the boundaries the human keyholder has set up in advance as free-text rules. It does not invent its own requirements.
The Model Context Protocol as the Link
The Model Context Protocol (MCP) is an open standard that lets AI assistants access external systems in a secure, structured way. You can think of MCP as a standardised port for AI applications — comparable to USB-C for devices.
The Chastity Tracker exposes an MCP server. An AI assistant such as Claude connects to this server as a client and gains two kinds of access:
- Reading: the current lock status, wear time, open inspections, training goals, the offense record and more.
- Writing (directives): setting lock periods, requesting inspections, adjusting training goals — all as clearly defined actions.
How a Decision Is Made
The flow is deliberately kept traceable:
- The AI keyholder reads the current session state via the MCP server.
- It compares that state against the human keyholder's free-text rules.
- Within those rules, it decides on a directive — for example extending a lock period or requesting an inspection.
- The action is executed and logged.
So the human keyholder defines the frame ("If a photo proof is late, extend by no more than 24 hours"), and the AI fills it in. It does not override those limits.
Everything Is Audited
Every directive the AI keyholder issues lands in the tracker's action log. This is not a side effect but a core principle: if you want to know later why a lock period was extended, you can trace it completely.
This transparency is also why an AI keyholder is not a trust risk compared to a purely human setup — provided both parties understand and accept the configured rules.
Mainly for Self-Hosting Setups
An honest assessment: the AI keyholder is not a "button for everyone". The MCP integration is most worthwhile for self-hosting setups, where you run your own tracker container and control the configuration. There you can expose the MCP server deliberately, connect an AI client of your choice, and write the rules in as much detail as you like.
For portal-only users the topic is less relevant — the portal is a free, friends-and-family service running on someone else's server and is not designed for this kind of individual integration.
No Hype, Just a Tool
An AI keyholder provides consistency — whether it takes routine work off a human keyholder or takes on the keyholder function for a sub without a partner. That is all it is, and all it is meant to be. In both cases the rules, and with them the responsibility, stay human: either the keyholder's or the ones the sub set for themselves. The AI is the assistant that enforces those rules reliably and transparently.
To see what this looks like in practice, read the example chats between keyholder and AI.