coreCerebrum

Agentic development workstation

The board is the interface.
Every card is a running agent.

Not an assistant bolted onto a task list. Each card spawns a real Claude Code session in its own git worktree, on its own branch, with memory that survives the session and the teammate.

~/.corecerebrum/worktrees · 4 lanes · live simulation
Backlog0
In flight0
Review0
Needs you0

Simulation of the board. Real captures from a running install further down.

The problem

You have outgrown a single session.

One agent, one window, one branch. You wait while it works. You lose the thread when the context fills. What it learned about your codebase yesterday is gone today, and nothing it learned reaches the person sitting next to you.

Without coreCerebrum

Serial work

One task at a time. The agent is the bottleneck, and you are idle while it thinks.

Amnesia

Every session rediscovers the same conventions, the same gotchas, the same decisions.

With coreCerebrum

Parallel lanes

Many sessions at once, each on its own branch. You review instead of waiting.

Shared memory

Decisions and gotchas persist, and they reach the whole team, not just you.

The biggest win

We do not tell you it is cheaper. We show you what it cost.

Every competitor in this category talks about speed. None of them shows you token economics at all. coreCerebrum reads Claude Code's own per-turn transcript, attributes real tokens to individual tool calls, and keeps the context lean enough that most of it never gets paid for twice.

/cost · session breakdown
input 2,189,207
cache_create 671,431,572
cache_read 37,337,552,731
output 68,049,861
paid equiv. 4,966,963,149
uncached would be 38,011,173,510
reduction 86.9%
Prompt cost reduction0%
Served from cache0%
Sessions measured0
Assistant turns0
Prompt tokens38.0B
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 holds between 86.6% and 88.0% across every TTL mix. This is a measurement, not a product benchmark, and we will not round it into a marketing number.

Where the waste goes

Agents burn tokens re-reading files they already read, re-deriving decisions that were settled last week, and hauling context that stopped being relevant twenty tool calls ago. That spend is invisible on every other board.

live session log
14:22:07 worktree created
14:22:09 reading src/auth/guard.ts
01

Re-read detection

The harness tracks what a session has already read and estimates the cost before it reads it again.

02

Memory recall

Past decisions and gotchas surface on the prompt that needs them, so they are not rediscovered from scratch.

03

Per-tool attribution

Real usage read from the transcript Claude Code actually sends. Not an estimate.

01

Parallel agent sessions

Every card runs a real Claude Code session. Not a queue, not a job runner. Many at once, each with its own context.

02

Worktree isolation

One git worktree per card, on its own branch. Agents cannot collide, and nothing lands on main by accident.

03

Persistent team memory

Decisions, conventions and gotchas outlive the session, and sync to the team rather than dying in one transcript.

04

Session control

Watch the diff, read the log, stop it, or take the wheel. The board is a control surface, not a status page.

Architecture

Isolation is the feature.

Parallelism without isolation is just a faster way to create merge conflicts. Every session gets a real worktree on a real branch, so ten agents working at once produce ten reviewable diffs, not one tangled working tree.

Full capability list
worktree/11fd0 fix/auth-guard worktree/9ab2c perf/cache-tenants worktree/4d7e1 feat/idempotency worktree/c03f8 chore/vitest The board 12 cards Memory shared, persistent team docs sync

Real interface

This is the product, not a rendering of it.

Captured from a running install. Card, project and session names are pseudonymized; the interface, the counts and the states are real.

board · lanes, branches, diffs, live session ticker
The coreCerebrum lane board: columns of cards showing repository, task, branch name and diff counts, with a review panel and a live session ticker along the bottom.
cross-project inbox
Cross-project inbox listing every card across all projects, with counts for needs pickup, in flight and review, and an active sessions rail.
work rail · live status
Work rail showing each item with a live status indicator: running, active, or needs you.

Underneath

Agents are only as good as the harness around them.

One shared configuration reaches every teammate identically: skills, commands, review agents, and a guard layer that refuses the operations you never want an agent to run.

Open the harness
Skills24
Commands22
Review agents5
Hooks10

GUARD

Destructive operations are denied, not warned about

rm -rf /, force-push to main, DROP TABLE, chmod 777, pipe-to-shell, and hardcoded secrets in writes.

CONTEXT

The bug log opens before the file does

Known failures and do-not-repeat entries for the exact file being edited, surfaced before the edit, plus a warning at three edits to one file.

ACCOUNTING

Token spend attributed per tool

Read from the authoritative per-turn transcript, so the number is what was actually sent, not a guess.

Honest positioning

Running agents in parallel is table stakes now.

Several tools do that, and some of them are free. The difference is what survives when the session ends, and whether anyone can tell you what the work cost.

CapabilityTypical agent boardcoreCerebrum
parallel sessionsYesYes
worktree isolationYesYes
memory across sessionsNot offeredPersistent
memory across the teamNot offeredSynced
token and cost accountingNot offeredPer tool, per turn
shared skills and commandsNot offered24 and 22
destructive-operation guardNot offeredHard deny
docs syncNot offeredIncluded

"Typical agent board" reflects the published feature and pricing pages of the parallel-agent tools we compared in August 2026. Their pages name parallel agents, workspaces, cloud hours and collaboration. None of them names persistent memory, team memory, cost accounting, or a guard layer.

Stop waiting on one agent.

$150 per user per month. Purchase, download, and run your first parallel session today.