How Many Magical Servitors Can You Work With? The Answer Is Not the One You Expected
Ask how many magical servitors you can work with at the same time and most answers give you a number, usually somewhere between seven and twelve, and none of them are defensible. The number people quote is a bookkeeping number. It answers a question you did not ask, because you were not asking about storage, you were asking about capacity. There are two ceilings and they are nowhere near each other. The first is how many specifications you can keep under review, which is a memory problem and which the servitor-craft literature actually addresses. The second is how many distinct states of attention you can enter, hold, and come back out of cleanly, and that is a nervous-system problem, it has almost nothing to do with magic, and it is the one that ends careers. One of those ceilings is generous and boring. The other is small and it moves.
Two different questions hide inside one sentence
Strip the question down and it splits cleanly in two. How many specifications can I maintain? That is a review problem, and its answer depends on whether you can answer three questions about each without looking anything up. How many can I be in contact with at a time? That is a state problem, and its answer depends on what happens to you when you try to be in two altered states at once.
Practitioners overwhelmingly answer the first question and call it the second. That is why the number everyone quotes is somewhere in the seven-to-twelve range. Seven to twelve is how many index cards a person can keep straight. It is not a number that came from anybody's nervous system.
Keeping the two apart matters practically. If you build to the bookkeeping ceiling, the failure is drift: your count stays stable while your usefulness falls, and you experience it as the practice getting boring. If you build to the state ceiling, the failure is worse than boring. It is the kind of failure that shows up as sleep disruption, agitation, difficulty being alone, or a sense of background pressure that does not lift when you stop. Those are the ones that end up in a doctor's office, and the doctor has no framework for them, so the cost lands entirely on you.
The bookkeeping ceiling, for completeness
You can keep about four specifications without an external system. Five to fifteen, you need a written record and a quarterly count. Above fifteen you are running something that has roles and a schedule and needs pruning, and the honest question stops being how many you have and becomes how many reviews you actually completed last quarter.
That is the whole answer to the first question. It is worth having and it is not what is interesting, so it gets one paragraph. The interesting answer is the next one, and it is smaller and stranger.
Worth noticing what the bookkeeping ceiling does not care about: how powerful a servitor is, how well it was charged, how long ago you built it, whether it corresponds to a Goetia spirit with thirty legions. None of that enters the calculation. It is a memory budget spent entirely on specifications.
The state ceiling is about as small as it sounds
Here is the number most practitioners would find uncomfortable: the real ceiling on active contact is one.
Not one at a time on different days. One at a time, in the sense that being in contact with a servitor is a state you enter, and a state has an entry cost, and the entry cost is a narrowing of attention. You cannot be in two narrowings at once. The second one is not a second contact, it is a thought about the first one, and the difference is not subtle once you know to look for it.
The error is easy to make and expensive. Someone builds a second servitor, sits down to meet it, and experiences a familiar sense of presence, and attributes it to the new construct. In a lot of cases they are correct. In a meaningful fraction they are meeting the first one, or their own expectation of a meeting, which is a well-practiced state in anyone who has done this for more than a month. The signal that this is what happened is that the new one feels reassuringly familiar within about ninety seconds.
A second, subtler version: the second servitor's specification gets quietly overwritten by the first one's. You go in asking about a money question and come out having received general advice about your life. The construct did not change. You did, and you did not notice, because the content was fluent.
What a contact actually costs, stated plainly
Entering a contact state is a physiological event, and the tradition that treats it as a purely symbolic exercise is doing itself a disservice. Here is what is happening: attention narrows, the ordinary ambient awareness of time and surroundings drops out, and if anything is moving in your environment you will not notice it. That is not mysticism, that is what narrowed attention is.
In a low-stakes setting this is unremarkable. Sitting at a desk for twenty minutes with a sigil is nobody's risk. In a setting where the stakes are, the same narrowing that produces the contact is what produces the failure, and the two are the same process. This is the single most useful thing to internalise about contact practice: the mechanism that makes the work work is the mechanism that makes it dangerous, and you cannot optimise one without touching the other.
The practical consequence: contact practice and sleep deprivation do not mix, and neither does contact practice and unresolved conflict, and neither does contact practice and a house with stairs in it. None of that is a spiritual caution. It is a description of what happens when narrowed attention meets an environment it is not tracking.
The accumulation question nobody asks properly
There is a difference between running many servitors and running many constructs. Most of the well-meant material on this point slides past it, and the slide is where the actual problem lives.
A specification is a written instruction you have made and can review. A construct is a presence in your attention. The first is inert until you enter a state, and reviewing it requires nothing from you. The second exists only during a state and can persist afterwards as an expectation. You can hold a hundred specifications and contact three of them a year and the other ninety-seven will never do anything at all, including possibly you.
The failure mode is that specs become constructs without anyone deciding it. You think about a servitor often enough over a few months and it starts to feel present, and the feeling is real as a feeling, and the question of whether the specification was ever correct stops being asked. What you now have is an expectation with a name attached, and expectations are much cheaper to acquire than specifications and much harder to review.
The distinguishing question is not how strongly you feel it. It is whether you can still state the specification, out loud, in the imperative, and whether you would be willing to retire it today. If both of those work, you have a construct with a specification attached, which is the good version. If the first works and the second does not, you have an expectation with a specification filed somewhere near it.
A protocol for finding your own ceiling
Do not start with a target. Start by finding where your first ceiling is, because for most people it is lower than they expect and finding it is the useful result.
The week one protocol: write one specification, something small and weekly, with a refusal clause and a review date. Contact it twice in the first week, on two separate days, and after each contact write one sentence about what you noticed, including the case where you noticed nothing. Do not build a second construct this week, whatever happens. The point of the week is the contact, not the count.
Week two: contact the same specification twice more. Add nothing. What you are looking for is whether the second contact was easier to enter, and if it was, whether the content got thinner. Ease of entry with thinner content is the normal trajectory and it is not a problem, it is what a specification does over time. The problem version is the same ease with content that gets more specific, because that usually means you are elaborating rather than receiving.
Week three: build the second specification. Contact the first twice and the second once, on separate days, never in the same sitting. After each, one sentence. If you cannot tell which construct you were in contact with in one of those three sessions, that is your answer for now and you do not need a third.
Where the review belongs, and why it is not optional
The review is the only mechanism in this entire structure that can tell you the truth, and it is the one everybody drops first. There is a reason, and it is a good one: building is satisfying and reviewing is not, so under any kind of pressure the satisfying one wins.
A quarterly ninety-second count is enough for the bookkeeping half. How many specifications exist. How many were actually performed. How many were retired. The middle number is the one that embarrasses people, and it is the one that tells you something real, because a specification that is never performed is a decision you made and then quietly reversed.
For the state half, the review is harder to automate and more useful. Once a month, ask three questions. Could I state each specification in the imperative without reading it. Could I say what each one refuses to do. Was there any week in the last month when I found it hard to be alone, or hard to settle, or where the background hum was louder than the day warranted. That third question is the physical one and it is the one that matters, and it is worth asking even if the answer is yes every month, because the trend is the information and a single month's reading is not.
If the answer to the third question is yes and reducing the count of contacts did not change it, that is not a servitor problem and you should stop treating it as one.
Signals, sorted by how much they should worry you
The benign ones, which mean nothing on their own: contact gets easier over months, the specifications get shorter as you learn what you actually need, you start skipping the ritual apparatus and just sit down, and you become more bored by the whole thing rather than less. Boredom in a contact practice at the eight-month mark is a sign it is working, not that it has failed.
The ambiguous ones: contact content that becomes more emotionally resonant over time, or that resolves more questions than the question you brought. This is the classic signature of a construct that has accumulated your expectations. It is not dangerous. It is not accurate either, and the correction is to go back to the specification, not to the contact.
The ones that should stop the practice this month, not next month: sleep that gets worse across a fortnight rather than one bad night, a background sense of pressure or being observed that does not resolve when you stop, agitation in the specific window after a sitting, or any physical symptom that reliably follows a practice and reliably does not follow an ordinary evening. That last one is a pattern, and patterns that follow a specific practice are data, not a test of your commitment. Reduce the frequency by half and see whether the symptom moves. If it does not move, the practice is not what is causing it and you should get it looked at anyway, because the symptom is real either way.
Five ways practitioners talk themselves into a bigger capacity than they have
Counting specifications and calling it capacity. The most common error, and the one that produces the seven-to-twelve figure, because it is a memory number dressed up as a limit.
Adding constructs during a period when you are depleted. The argument for restraint is not mystical, it is that the narrowing that produces contact is the narrowing that removes your capacity to notice you are depleted, so a depleted practitioner is the one least able to self-police frequency. This is the same mechanism as the one that makes contact in a cluttered room a bad idea, and it is why the rule is to add on the strength of a good month rather than on a good week.
Treating a strong feeling as proof of a construct. Intensity is a reasonable indicator of attention and a terrible indicator of source. A construct you have been thinking about daily for a year will feel more present than one you charged beautifully in March and have not touched since, and it will feel more present for the boring reason that you have been rehearsing the expectation.
Confusing persistence with function. A construct that keeps surfacing unbidden is not a well-charged one. It is one that has become an expectation, and persistence is a symptom of that rather than evidence of efficacy.
Escalating on a good streak. The most reliable accounts of people who got into trouble with this are accounts of someone whose practice went well for two months and then became the only thing they did, and the mechanism is not mysterious: the practice worked for them often enough that it became their primary strategy for handling difficulty, which is exactly the shape of a thing that eventually runs a person into the ground.
What the question is really asking
Strip the framing question out and the real question underneath is usually not about capacity at all. It is something closer to whether there is a point at which doing a lot of this becomes a different kind of life. Somebody asking how many servitors they can run is very often somebody who has noticed that the count is their main source of a feeling they like, and would like to know whether there is a ceiling before they find out the expensive way.
There is a ceiling, and it is low, and it is lower than the question implies. The reason the literature does not engage with this properly is that the numbers people circulate are answering a different question, and a book that answered the real one would have to say that the practice is structured so that most of it happens away from the practitioner, which is the opposite of what the person asking wants to hear.
The version of this that is actually true: the specifications are the valuable part and they are cheap. You can hold dozens and touch four a year. The contacts are the valuable part and they are expensive and there is one at a time. Everything else, the entities, the charge, the correspondences, the number of legs, is a way of filling in the parts of the story where the actual work happens, which is in the room, alone, with your attention narrowed, doing one thing.
Frequently Asked Questions
How many magical servitors can I work with at the same time?
One at a time in the active sense. A contact state is a narrowing of attention and you cannot be in two narrowings at once, so the second contact is a thought about the first rather than a second contact. If by working with them you mean maintaining them, the bookkeeping limit is roughly four without a written record, five to fifteen with one, and above fifteen you are running something with roles and a pruning schedule rather than a practice. Those are two different ceilings and mixing them up is how people end up with twenty specifications and no way to tell whether any of them does anything.
Is it bad to have a servitor I think about all the time?
Thinking about a construct frequently is the ordinary way a construct becomes an expectation instead of a specification, and an expectation feels real as a feeling while carrying none of the information a specification carries. The test is not how strongly you feel it. It is whether you can still state the specification out loud in the imperative, say what it refuses to do, and would be willing to retire it today. If those three work, what you have is fine. If the first works and the retirement answer does not, the feeling is telling you about your attachment rather than about the construct.
Why do I feel more presence with a servitor I built early and barely use?
Because you have been rehearsing the expectation rather than maintaining the specification. A construct you have thought about daily for a year will feel more present than one you charged carefully in March and have not touched since, and the reason is the least interesting one available: you have been thinking about it. Intensity of feeling is a decent indicator of attention and a poor indicator of source, which is why it should never be used to adjudicate which of two constructs a contact was with.
Can servitors affect my sleep or make me anxious?
Contact practice narrows attention, and that is the same process that produces the work and the risk, so it is not surprising that it can disturb sleep or produce agitation, particularly in someone who is already depleted. What is worth acting on is a pattern rather than a single bad night: sleep getting worse across a fortnight, background pressure that does not lift when you stop, or a physical symptom that reliably follows a sitting and reliably does not follow an ordinary evening. Halve the frequency and see whether the symptom moves with it. If it does not, the practice is not what is causing it, and the symptom is real and worth getting looked at regardless.
Do I need to keep every servitor I have built?
No, and a high retirement rate is the healthy case. A standing instruction you have decided not to maintain is a small open loop that will not close itself, and most of the value of a specification is that it corresponds to something you actually do. The quarterly ninety-second count is enough: how many exist, how many were performed, how many were retired. The middle number is the one that tells you something, and it is usually lower than expected.