API & Data Platform
XeroML is built on open data principles. Everything in the platform is accessible via API, and your data is always exportable in standard formats. This makes it straightforward to build custom workflows, feed data into external analytics tools, or use XeroML as the data layer for your own LLM tooling.
What You Can Build
Cost management Implement billing or usage-based pricing by reading per-user or per-project token consumption from XeroML’s API.
Analytics integration Feed evaluation scores and trace metadata into your existing reporting tools — PostHog, Mixpanel, Looker, or any data warehouse.
Model improvement Export trace data (inputs, outputs, scores) to build fine-tuning datasets for your own model training pipelines.
Behavioral analysis Join LLM evaluation outcomes with user behavior data to understand how AI quality affects business metrics.
Custom dashboards Query XeroML’s metrics API to build domain-specific dashboards beyond what the built-in UI provides.
Access Methods
| Method | Best for |
|---|---|
| MCP Server | Connecting AI coding agents to your XeroML data |
| Public API | Programmatic read/write access from any HTTP client |
| Query via SDK | Fetching traces, scores, and datasets from Python/TypeScript |
| Export from UI | One-time or ad-hoc CSV/JSON exports |
| Blob Storage Export | Continuous export to S3, GCS, or Azure Blob |
Authentication
All API access uses Basic authentication with your XeroML API keys:
# Public key as username, secret key as passwordcurl -u "pk-xm-...:sk-xm-..." https://cloud.xeroml.com/api/public/v2/tracesAPI keys are scoped to a project and available in Project Settings → API Keys.
API Reference
The full OpenAPI specification is available at:
- Cloud:
https://cloud.xeroml.com/api/public/v2/openapi.json - Self-hosted:
https://your-xeroml-instance/api/public/v2/openapi.json