coreCerebrum

Pricing

$150 / user / month

One tier. Every capability, the full harness, memory and docs sync, and the cost accounting. There is no cut-down plan, because a cut-down version of this is not worth running.

Purchase, then download. Model usage is billed by your own Claude subscription or API account, which means you see and control that spend directly.

Includedall of it
parallel sessionsunlimited
worktree isolationincluded
persistent memoryincluded
team memory syncincluded
docs syncincluded
cost accountingincluded
harness: 24 skills, 22 commandsincluded
guard layer: 10 hooksincluded
review agents5
cowork surfaceincluded
custom skills, commands, hooksyours to write
Some tools in this category are free or close to it. That is real, and you should weigh it. What you are buying here is the layer above running agents in parallel: memory that persists and reaches your team, a guard layer that refuses the operations you cannot undo, a shared skill and command set, and accounting that tells you what the work cost. Those are the parts nobody else ships.
CapabilityFree and low-cost toolscoreCerebrum
run agents in parallelYesYes
isolated worktreesYesYes
memory across sessionsNot offeredIncluded
memory across the teamNot offeredIncluded
token and cost accountingNot offeredPer tool, per turn
shared skills and commandsNot offered24 and 22
destructive-operation guardNot offeredHard deny
docs syncNot offeredIncluded

Comparison reflects the published feature and pricing pages of parallel-agent tools reviewed in August 2026. If one of them ships these, this table should change, and we will change it.

The gap

Parallel agents are table stakes. What survives the session is not.

The first time a tool runs six agents at once it feels like magic. The second week you notice every one of them relearning the same thing, and nobody can tell you what any of it cost. That is the problem this price is against.

Cost accounting

The seat is not the only line on the invoice.

Model usage is the other one, and on a team running agents all day it is the bigger one. Keeping context lean is not a nice-to-have at that scale, it is the difference between a tool you can afford to leave running and one you cannot.

Prompt cost reduction0%
Served from cache0%
Sessions measured0
Assistant turns0
Measured on one machine across 1,618 Claude Code sessions and 105,724 assistant turns, priced at published cache multipliers: reads at 0.1x base input, writes at 2x for one-hour TTL and 1.25x for five-minute. The figure sits between 86.6% and 88.0% across every TTL mix. It is a measurement of one installation, not a guarantee about yours, and your own number will be visible to you from the first week.

SEATS

One seat per person

$150 per user per month. Anyone who runs sessions needs a seat.

SHARED

Memory compounds

The value of the memory layer goes up with each person on it, because what one session learns the next teammate starts with.

LARGER

Bigger teams

If you are rolling this out beyond a handful of people, talk to us about it directly rather than buying seats one at a time.

Is there a free trial?

No. You purchase, download, and run it against your real repository. If it is not right for you, tell us.

Do I need my own Claude subscription?

Yes. Sessions run against your account, so model spend is yours, visible to you, and under your control.

Where does my code go?

The board runs locally against your machine and your checkouts. Worktrees are directories on your disk.

Is the harness extra?

No. Skills, commands, review agents and the guard layer are included, and you can write your own.

How many agents can I run at once?

As many as your machine and your model account will carry. The board does not cap it.

What if I want to leave?

Your work is git branches and your memory is files. Both stay where they are.

$150 per user per month.

Purchase, download, and run your first parallel session today.