spaturzu is an SDK designed for attributing large language model (LLM) costs to individual AI agents, runs, and projects within applications. It addresses the challenge of understanding which specific agents or features are responsible for LLM usage costs, especially when multiple agents share the same provider key and the provider’s invoice aggregates all expenses into a single figure. The tool is intended for teams and developers deploying multiple AI agents and seeking detailed cost breakdowns for better financial oversight and accountability.
The SDK integrates directly into application code and operates in-process, meaning it does not act as a proxy and does not require any changes to prompt structures. It supports both Node and Python environments and works with several LLM providers, including OpenAI, Anthropic, Bedrock, Gemini, and Mistral. By tagging each LLM call with the relevant agent, run, and project, spaturzu records token counts and associated costs at the moment the request is made. This enables users to see cost visibility shaped by their code, with per-agent attribution and 7-day or 30-day cost rollups available. The platform provides run-level traces, allowing users to drill down from expensive runs to the specific provider calls that contributed to those costs.
spaturzu offers features such as daily or monthly budget caps per project and can trigger alerts via webhooks or Slack when spending thresholds are crossed, helping teams avoid unexpected overages. A key focus of the tool is privacy: prompt content and model responses never leave the user’s servers, as only token counts and computed costs are sent to spaturzu. This privacy-by-design approach makes it suitable for organizations with regulatory or compliance requirements, such as GDPR, HIPAA, or SOC 2.
Installation involves adding the SDK as a dependency and swapping a single import in the codebase. No additional infrastructure is required. The service is available for free, with account creation occurring automatically upon first use with an email sign-in. spaturzu is positioned as a per-agent LLM cost attribution tool, distinguishing itself from other LLM observability platforms by focusing specifically on cost breakdowns without routing prompt data through external systems.
In the LLM eval & observability space, spaturzu takes a focused approach. Allowing developers to attribute and monitor LLM usage costs per agent in Python applications. spaturzu is an open-source project aimed at developers using LLMs in Python. The project is open source (MIT). It runs on the web and the command line.
spaturzu first shipped in 2026. The project is developed in the open on GitHub with 17 commits in the last 90 days. Among its 4 catalogued features are cost-attribution, llm-monitoring, and python-integration.
Latest indexed changes and source events
Other apps tracked under the same category.
