Cost estimation

Estimating token and API costs before tasks

• A Preflight Registry for File Hashes and Token Counts
• Budgeting a Full Mutation Run
• Costing a Caller Inventory Pass Across a Monorepo
• Costing a Retrieval Eval: Golden Queries and Judge Calls
• Costing a Self-Review Gate: Tokens, Minutes and Review Time
• Costing a Wrong Tool Pick in Tokens and Retries
• Estimating a Dependency Change Before an Agent Installs
• Estimating Token Cost Before You Send an Agent a Task
• All of Cost estimation

Agent budgeting

Budgeting agent runs and workflows

• Blocking a Run When the Budget Check Fails: Exit Codes and Logs
• Enforcing Budgets Across Child Agents and Sub-Tasks
• Reserving Rate-Limit Capacity Alongside Token Spend
• Setting a Maximum Review Cost per Agent Branch: A Policy Guide
• Setting a Per-Task Budget an Agent Cannot Exceed
• A Budget Ledger Schema: Reservations, Refunds, Charges
• A Fallback Path When the Reservation Broker Is Down
• Budgeting a Shadow Copy of Production Traffic
• All of Agent budgeting

Spend tracking

Tracking and reconciling AI model spend

• Measuring Token Spend from Context Files and Skills
• Proxying Reservations for Sub-Agents on an Agent Platform
• A Broker Log Schema That Joins Reservations to Spend
• A Budget Dashboard: Traces, Spend and Stop Reasons Joined
• A Check Cost Ledger: Tokens, Time and Spend in One Table
• A Rejected Count: Logging the Preflight to Provider Gap
• A Shadow-Run Log Schema: Fields for Comparing Model Outputs
• How to Log an Agent Run So You Can Tell Where It Went Over
• All of Spend tracking

Cost optimization

Caching and routing to reduce agent costs

• Choosing a Model Per Task from Pass Rates and Token Counts
• Routing Only the Hard Cases to the Expensive Model
• Testing Output Against a New Model Without Burning the Budget
• Caching and Routing: When a Cache Hit Should Still Escalate
• Hashing a Prompt Template into the Same Key
• Pricing a Cache Hit and a Cache Miss in Dollars
• Prompt Caching for Agents: What to Cache and Skip
• Routing Agent Requests to the Cheapest Model That Passes Evals
• All of Cost optimization

Budget overruns

Handling budget overruns and orphaned agents

• A Grace Allowance When the Quota Service Is Down
• Fan-Out Without Fan-In: Handling Orphaned Sub-Agents
• Killing a Run Mid-Stream Without Corrupting State
• Postmortems for Budget Overruns: Writing the Run Record
• Reading the Override Log for Patterns: Queries and Thresholds
• Refunding Reserved Capacity When an Agent Task Fails
• Reporting Agent Overages Up: Metrics a Manager Can Act On
• Reserving Budget for Orphaned Sub-Agents
• All of Budget overruns