Skip to main content

Foxhound

Open-source observability for AI agent fleets.

Foxhound gives you deep visibility into every AI agent call — traces, evals, cost, latency, and policy violations — so you can ship AI safely at scale.

Explore the docs

SectionDescription
Getting StartedInstall Foxhound and send your first trace in minutes
TypeScript SDKFull API reference for the Node.js / TypeScript SDK
Python SDKFull API reference for the Python SDK
IntegrationsDrop-in wrappers for LangGraph, CrewAI, Mastra, and more
MCP ServerUse Foxhound tools from any MCP-compatible AI assistant
Prompt ManagementVersioned prompt registry with labels, caching, and trace linking
CI/CD Quality GateBlock deploys when eval scores regress
Evaluation CookbookRecipes for scoring, judging, and curating eval datasets

Quick install

# TypeScript / Node.js
npm install @foxhound-ai/sdk

# Python
pip install foxhound-ai

Why Foxhound?

  • Full-trace observability — every LLM call, tool invocation, and agent hop captured automatically
  • Policy enforcement — detect PII leakage, prompt injection, and off-topic responses in real time
  • Eval pipelines — score outputs with LLM-as-a-judge or human review, then gate deploys on those scores
  • OpenTelemetry native — works with your existing OTel stack; no lock-in
  • Audit-ready — structured logging of every agent action for review and debugging