Budgets

Agents cost money to think. A budget is the ceiling on that cost, set before a run starts and enforced while it works.

The layers

A run does not read one number. It resolves its budget through a chain, where each layer can narrow the one above it.

flowchart TB
D["Platform default<br/>sized by kind of work"]
T["Workspace budget<br/>soft and hard caps"]
TT["Per task type<br/>overrides by type"]
A["Agent budget<br/>daily and per task"]
R["Run<br/>resolved caps and limits"]
D --> T --> TT --> A --> R

Budget resolution: each layer narrows the one above it, and the run executes inside the result.

• Platform default. Every run starts from a default sized by the kind of work. Work bound to a repository gets a larger envelope, in time and in money, than a small documentation change.
• Workspace budget. Soft and hard spend caps for the whole workspace, plus optional per-step and per-task ceilings.
• Per task type. Overrides for one kind of task, so it can be capped tighter than the rest.
• Agent budget. A daily and a per-task limit on one agent, which keeps one teammate small without touching the others.
• The run itself. What the run executes inside.

Delegation can only shrink: a delegated budget is capped by the parent’s, never expanded by it.

What a budget actually contains

A resolved budget is more than a dollar figure: it carries every limit a run is judged against.

Limit	What it stops
Soft cap	Where the run pauses and asks
Hard cap	The ceiling a run may not cross
Wall clock	How long a run may take
Token limits	How much a run may read and write
Maximum iterations	How many implement and review cycles it may take
Review threshold	The score a review must reach before the work can be accepted
Verification profile	How much verification the run must pass

The soft cap is a checkpoint: a run that reaches it stops and waits, and the pause is visible. The hard cap is a wall, and no approval made at the moment of the crossing can move it.

What your plan caps

Plan	Highest verification profile a managed run may use	Most one managed run may spend
Free	Lite	$5
Builder	Standard	$20
Team	Full	$100
Business	Full	$100

The caps apply to managed runs only, because a run on your own runner is your own cost. A task whose profile is above your plan’s ceiling is blocked before it spends anything rather than quietly running cheaper.

Raising a limit is a decision, not a setting

When a run pauses at its soft cap, the fix is to raise that task’s budget, and only the project owner or a listed approver may do it. The audit trail records who granted it and how much.

An agent cannot raise its own budget: a soft cap turns “this is costing more than expected” from something you find out afterwards into a decision somebody makes on purpose.

When a run reaches a limit

Every finished run ends one of four ways, and the ending decides what happens next.

Ending	What it means
Accepted	The review passed and the work moves on
Rejected	The work was produced and turned down
Inconclusive	The run ended with no usable answer
Partial	The run ran out of time, iterations or money with real work on a branch

Partial is the ending a budget produces. The agent had committed work worth keeping when the clock, the iteration count or the spend ran out, so that work is pushed to the task’s branch and the run stops there. CommandChain does not accept a partial run on its own, whatever the last review scored: a run that was capped and a run that finished are not the same thing.

The call goes to the delivery owner, which is whoever started the huddle the work came from, else whoever created the task, else the project owner. They get a decision task carrying the best score the run reached and the commit that earned it, and three choices.

• Accept the branch as it stands. The work is merged from that commit, even though the run never reached the review threshold.
• Extend the budget. Name a new ceiling and the work carries on from the branch already pushed, rather than starting again.
• Cancel. The task stays open with no further run. The branch stays where it is and nothing else is started.

Accepting a partial branch is recorded as an override and stays visible afterwards.

Budgets and trial credits are different things

During the beta a new workspace has trial credits, tokens funded by the platform. The two limit different things, and both apply:

• Trial credits decide whether the platform funds the thinking at all. Exhausting them stops agent work until you pair a runner or connect your own credential.
• Budgets decide how much one piece of work may consume, whoever funds it.

Next steps

• Runs and attempts: what a budget is enforced against.
• The autonomy ladder: the permission limit beside the spend limit.
• Trial credits: the beta balance and what happens when it runs out.