Skip to main content
HoneyHive provides distributed tracing for AI applications, everything from RAG to multi-agent systems, giving you a hierarchical view of execution across agent invocations, LLM requests, tool calls, and handoffs.
Tree structure of a trace showing nested events

​
Getting Started

​
Automatic Tracing

The fastest way to start is automatic tracing, which instruments major LLM providers and vector databases with minimal setup using OpenTelemetry semantic conventions.
Using an agent framework? We currently document framework-specific setup for AWS Strands, Google ADK, and PydanticAI.

Quickstart

Begin tracing OpenAI requests with HoneyHive in minutes

Supported Integrations

Pre-built integrations for common model providers and frameworks
Deploying in Lambda, FastAPI, Flask, or Django? Read Tracer Initialization next to choose the right runtime setup pattern before you wire tracing into production.

​
Custom Spans

Automatic tracing covers LLM and vector DB calls. For any other function in your codebase (preprocessing, postprocessing, business logic), use the @trace() decorator to create custom spans that appear in your trace tree.

Custom Spans

Trace any function in Python/TS using decorators

​
Advanced Features

Distributed Tracing

Propagate trace context across service boundaries so multi-service calls appear in a single session

Sampling

Control which requests get traced in high-volume applications

Span Filtering

Drop noisy framework spans using prefix-based rules

Multi-Instance Tracing

Run multiple tracer instances for multi-tenant, A/B testing, or environment routing

Multi-Modal Tracing

Trace pipelines that combine text, image, audio, or video operations

Multi-Threaded Tracing

Propagate context across threads in Python applications

Tracing via API

Log events directly via REST API for languages without SDK support