Agentrics
Track, analyze, and reduce AI agent API costs across OpenAI, Anthropic, and Google — with a single SDK integration.
How it works
Agentrics wraps your existing AI clients (OpenAI, Anthropic) with a thin tracking layer. Every API call is automatically intercepted, costs are calculated server-side using live pricing, and the data is stored in your Agentrics account — scoped to the API key used.
- Generate an API key from Settings → API Keys
- Install the Python or JavaScript SDK and wrap your AI client
- View cost breakdowns by agent, model, and time period in the Dashboard
Supported providers
OpenAI
gpt-4o, gpt-4o-mini, gpt-4, gpt-3.5-turbo, o1, o3…
Anthropic
claude-3-5-sonnet, claude-3-haiku, claude-opus…
gemini-1.5-pro, gemini-1.5-flash, gemini-2.0-flash…
Key concepts
- Agent
- A named component of your system that makes AI calls. Assign each agent an
agent_id/agentIdto track it separately in the dashboard. Use a consistent naming convention likesummarizerorsupport-classifier. - Execution
- A single API call tracked by Agentrics. Each execution records the model, input/output token counts, calculated cost in cents, latency, and success status.
- API Key
- A secret key prefixed with
agntcs_sk_used to authenticate tracking requests. Each key is scoped to your user account — all executions tracked with a key appear in your dashboard.