Agent Cost Optimizer

Real-time agent operation cost tracking and autonomous budget allocation with cost-ceiling enforcement.

Features

• Real-time Cost Tracking: Track token usage across different LLM providers (OpenAI, Anthropic, Google).
• Budget Management: Set daily, per-session, and monthly budget limits for multiple agents.
• Ceiling Enforcement: Automatically check and block operations that exceed budget limits.
• Cost Analysis: Generate usage summaries and export data to CSV or Markdown.
• Alerting System: Get notified when agents reach specific budget thresholds (50%, 80%, 90%, 100%).

Installation

bun install

Usage

Record Token Usage

bun src/index.ts record -a agent-1 -s session-123 -m gpt-4o -p 500 -c 200

Check Budget Status

bun src/index.ts check -a agent-1 -s session-123

Set Agent Budgets

bun src/index.ts set-budget -a agent-1 -d 5.0 -s 0.5 -m 50.0

View Usage Summary

bun src/index.ts summary -a agent-1

List All Budgets

bun src/index.ts list-budgets

Architecture

• CostTracker: Manages the persistence and aggregation of cost records.
• BudgetManager: Handles agent-specific budget configurations.
• CeilingEnforcer: Core logic for determining if an operation is allowed based on current usage and budget.
• AlertManager: Monitors usage levels and triggers notifications at configurable thresholds.
• DataExporter: Utility for exporting cost data to various formats.

Configuration

Pricing data is maintained in src/providers.ts and can be updated to reflect current API costs. Data is stored locally in the data/ directory.

License

MIT