Development

Budgets and spend control for agents

• A Spend Cap Across Sub-Agents: When One Run Fans Out to Five
• A Token Budget That Leaves Room for the Model to Answer
• A Tool Call Budget per Run: Counting Writes Before the Start
• Caps on an Agent Finance Tool: Per Run, Customer and Day
• Setting Output Token Limits per Task Without Starving the Model
• A Cheap Model or a Larger One: The Break-Even Point
• A Child Run's Spend Attribution: Fields for Parent Aggregation
• A Context Budget Report for Every Agent Run
• A Cost Ceiling for CI Model Calls: Budgets and Hard Stops
• A Cost Ceiling for Model Calls: Budgets, Alerts and Hard Stops
• A Cost Ceiling per Event: Budgets, Alerts and Hard Stops
• A Cost-Aware Stop: Budgets That Fire Before the Next Model Call
• A Per-Fault Cost Table: Pricing Each Rejection Type
• A Refund Path for a Spend Overrun: Compensating a Long Child
• A Run-Level Spend Cap When One Task Fans Out
• A Token Budget for a Conversation: Tracking Spend Across Turns
• A Token Budget per Customer: Alerts and Hard Stops
• A Write Budget Across Runs: Daily Caps per Customer
• A Write Budget per Customer and per Day
• Alerts Before the Cap: Warning Thresholds That Leave Time to Act
• Caching Model Responses: Exact, Semantic and Prompt-Prefix Caches
• Caching Model Responses: Keys, TTLs and Invalidation
• Cost Per Resolved Request When a Person Handles It
• Fraud Signals on Agent Credits: Velocity and Repeats
• Propagating a Budget Through Sub-Agents Without Double Counting
• Reasoning Tokens Are Not Output Tokens: Budgeting CoT
• Reconciling a Counter After a Crash: Recovery Before the Next Run
• Reconciling Counters After a Crash: Spend Recovery
• Routing a Sub-Agent to a Cheaper Model on a Small Reservation
• Stopping Agent Retry Loops Before They Run Up Spend
• Stubbing Clocks and Headers So a Refusal Path Runs in CI
• Trimming History Without Losing the Thread
• When a Write Budget Fires: What the Agent Says
• When to Summarize History Instead of Dropping It