dexcost Documentation
Agent Unit Economics platform -- track LLM costs, non-LLM service fees, and retry cost attributed to customers, projects, and workflows.
Welcome to dexcost
dexcost is the Agent Unit Economics platform. It tracks every cost your AI agents generate -- LLM calls, API fees, tool calls, compute, and retries -- and attributes them to customers, projects, and workflows.
What makes dexcost different
- Full-stack cost tracking -- not just LLM tokens, but the API fees, tool calls, and compute that make up 27% of agent costs
- Retry cost as a first-class metric -- see exactly how much you spend on failed retries and where they happen
- Customer-level P&L -- know which customers are profitable and which are bleeding you
- Python SDK today, more on the roadmap -- v0.1.0 ships Python with auto-instrumentation for OpenAI, Anthropic, LiteLLM, Gemini, Bedrock, and Cohere. TypeScript, Go, and Rust SDKs are in active development.
Quick links
- Scan your codebase -- find untracked cost points in 30 seconds
- Quickstart -- start tracking costs in 5 minutes
- Python SDK -- install and instrument your first agent
- API Reference -- REST API for analytics and ingestion