Cognition Adds Hierarchical Budgets to Devin
The AI coding agent maker is letting companies cap and distribute Devin spending across business units, teams and projects.
Updated on Nov. 4, 2026 in Budgeting
Cognition has introduced hierarchical budgets for its Devin AI coding agent, allowing administrators to assign spending limits to business units, teams and projects. The feature lets central IT set limits without approving each individual request.
Why it matters
The feature followed customer feedback that single-team controls did not scale as Devin moved into broader organisational deployments. Budget controls give enterprises a way to manage the cost of AI agent usage across large engineering departments.
Budgets can be nested three levels deep, and separate spending controls apply to Devin running in the cloud and on engineers' desktops. Cognition has not disclosed how the controls will be priced or how widely they have been adopted.
The players
Cognition
Cognition is the AI company that develops Devin, an autonomous AI software engineering agent.
The details
Administrators can start with an overall budget and distribute it across business units, which can then allocate funds to teams and projects. Engineers receive homepage alerts when nearing a limit and can request more Agent Compute Units within the product, while administrators can enable efficiency-based approvals that raise limits up to a set ceiling. If a budget runs out mid-task, the Devin session stops.
Timeline
October: Cognition introduced the Efficiency Score, which estimates productive engineering hours relative to Devin consumption.
November 4, 2026: Article published on the hierarchical budget feature.
Market Dynamics
The budget controls extend Devin's Efficiency Score, introduced in October, which measures productive engineering hours against Devin consumption. Adding spending hierarchy moves Devin toward the enterprise cost-governance tooling that larger software deployments typically require.
Engineering teams using Devin can set spending caps by team or project and receive alerts before limits are reached. Engineers who run out of allocated compute can request more from the budget owner inside the product.
The takeaway
Organisations adopting AI coding agents can use nested budgets to keep spending aligned with team ownership. Setting clear limits and approval paths early helps avoid unplanned costs as usage scales.
Further reading
For broader context on managing spending, explore Budgeting guides on the site.
