# Currai

> Currai is observability for LLM applications. It traces prompts, completions, tool calls, token usage, latency, cost, sessions, users, and nested spans so teams can debug and evaluate production AI systems.

Currai is built for developers and teams shipping LLM products. It provides hosted ingestion, storage, and dashboards with first-party Python and TypeScript SDKs, Langfuse-compatible SDK surfaces, and OpenTelemetry OTLP ingestion.

## Core Product

- Product: Currai
- Category: LLM observability, AI observability, LLM tracing, prompt tracing, token cost tracking
- Website: https://www.currai.app/
- Pricing: https://www.currai.app/pricing
- Agent-readable pricing: https://www.currai.app/pricing.md
- Documentation: https://www.currai.app/docs
- API reference: https://www.currai.app/api-reference
- OpenAPI spec: https://www.currai.app/openapi.yaml
- Blog: https://www.currai.app/blog
- Support: support@currai.app

## What Currai Does

Currai captures one logical LLM operation as a trace. Inside a trace, model calls are generations, non-model work such as retrieval and tool calls are spans, and point-in-time markers are events. This structure lets teams inspect the full tree behind an LLM response instead of relying on app logs or screenshots.

Currai tracks token usage, cost, latency, user IDs, session IDs, environments, tags, inputs, outputs, nested spans, and errors. Teams use it to debug bad answers, inspect slow RAG pipelines, understand expensive users or sessions, and build production trace datasets for evaluation.

## Key Pages

- Introduction: https://www.currai.app/docs
- Installation: https://www.currai.app/docs/installation
- Authentication: https://www.currai.app/docs/authentication
- First trace: https://www.currai.app/docs/first-trace
- Troubleshooting: https://www.currai.app/docs/troubleshooting
- Generations: https://www.currai.app/docs/generations
- Nested traces: https://www.currai.app/docs/nested-traces
- Sessions and users: https://www.currai.app/docs/sessions-and-users
- Cost and tokens: https://www.currai.app/docs/cost-and-tokens
- OpenTelemetry: https://www.currai.app/docs/opentelemetry
- Prompts: https://www.currai.app/docs/prompts

## Product Capabilities

- Trace LLM app behavior with Python or TypeScript SDKs.
- Send OpenTelemetry spans to Currai through OTLP endpoints.
- Migrate Langfuse SDK instrumentation by changing the host URL.
- View full traces with generations, spans, tool calls, inputs, outputs, and errors.
- Roll up tokens, cost, and latency by trace, model, user, and day.
- Group multi-turn conversations by session and end user.
- Run prompt evaluations and A/B test prompt versions in production.
- Use hosted ingestion, storage, and dashboards without running a collector or database.

## Comparison Context

Currai is relevant for queries about LLM observability, LLM tracing, Langfuse alternatives, OpenTelemetry for LLM apps, prompt A/B testing, token cost tracking, AI agent observability, RAG debugging, and production AI monitoring.

## Freshness

Last updated: 2026-06-17