What problem does it solve?
Resource-heavy agent workflows need predictable limits so API rate, storage capacity, and connection concurrency don’t overwhelm downstream services or destabilize your deployment.
Core Features & Use Cases
- Quota configuration in golem.yaml: Define rate, capacity, and concurrency limits per resource (with environment scoping) and choose enforcement actions like reject or throttle.
- QuotaToken reservations and accounting: Acquire quota tokens once and use
withReservation to reserve expected usage and commit actual consumption (including variable-cost workloads like LLM token usage).
- Agent-to-agent budget splitting: Split a parent token to child agents for controlled delegation and merge returned tokens to restore remaining budget.
- Runtime quota updates: Adjust resource limits via the CLI so changes take effect for running agents.
Quick Start
Use the golem-quota-ts skill to add a quota definition for your target resource in golem.yaml, acquire the QuotaToken in your agent, and wrap each expensive operation with withReservation to enforce consistent limits.