Skip to content
Early access
Menu
Appearance

Pricing

One plan. No meters.

One price, every feature, however many agents you run. The model calls go out on your own Claude subscription, so the only thing that scales with usage is already yours.

plan1 of 1
$39USD per month

$32 per month, billed annually

Get early access

14-day trial, no card required

macOS 14 or later, coming soonWindows and Linux, private beta

It runs on your Claude subscription.

costper month6 rows · 4 at zero
  1. A sixth agent, then a tenthno change
  2. A second project, then tenno change
  3. Every work item the board opensno change
  4. Every event it ever appendsno change
  5. The model calls your agents makeyour Claude planbilled on the login you already use, never by us
  6. A teammate on their own machinetheir own subscription

The product runs on your machine, so there is no per-agent cost for us to pass on. The one thing that scales is the model call, and it goes to the Claude subscription you already pay for.

Straight answers.

Does running more agents cost more?

Not from us. The number on this page does not move with agents, projects, work items or events. Model calls run on your own Claude subscription and are billed there, not here.

Which Claude plan do I need?

The same one Claude Code needs. It signs in with the login you already use, so if the CLI runs today, this runs today.

Which coding agents does it run?

Claude Code today, through your own login. The runner sits behind an adapter seam, so a new agent is an adapter, not a rewrite.

What does it not do?

It does not write code (the agents do), host anything, or replace review. It makes review possible at fleet scale by keeping the record review needs.

Why is it not open source?

It is the product, and the subscription funds building it properly. Your data stays local and in an open format regardless.

What happens to my data if you go away?

Nothing. The graph is a single file on your machine in an open format that standard database tooling reads, so it outlives us; nothing was ever held anywhere else.

How do I know the agents followed the rules?

You check the trail. Every gate decision appends an event naming the rule and the diff, and the breakers bind at write time, not by asking agents to behave.

Can my team share a board?

Not yet. Shared boards mean a server, which is a product decision we have not made, rather than a feature we forgot.