AgentCost provides comprehensive spend analysis for AI agents, specifically targeting Claude Code session logs. It functions as an MCP server, allowing users to gain insights into their token usage, identify inefficiencies, and find opportunities for optimization. A free preview tier offers basic cost breakdowns, while paid tiers provide advanced features like waste pattern detection, per-tool cost attribution, and actionable recommendations, all handled via x402 micropayments without the need for API keys or subscriptions.
Key Features
01Tool usage frequency and top expensive turns
02Waste pattern detection and optimization recommendations (paid)
03Micropayment-based access via x402 on Base mainnet
04Detailed token breakdown (input/output/cache)
05Free preview tier for basic cost analysis
060 GitHub stars
