Foxhound
Compliance-grade 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
| Section | Description |
|---|---|
| Getting Started | Install Foxhound and send your first trace in minutes |
| TypeScript SDK | Full API reference for the Node.js / TypeScript SDK |
| Python SDK | Full API reference for the Python SDK |
| Integrations | Drop-in wrappers for LangGraph, CrewAI, Mastra, and more |
| MCP Server | Use Foxhound tools from any MCP-compatible AI assistant |
| Prompt Management | Versioned prompt registry with labels, caching, and trace linking |
| CI/CD Quality Gate | Block deploys when eval scores regress |
| Evaluation Cookbook | Recipes for scoring, judging, and curating eval datasets |
Live sandbox
Explore Foxhound without setting up infrastructure. The sandbox ships with 568 seeded traces across a realistic seven-day operating story:
git clone https://github.com/caleb-love/foxhound.git
cd foxhound && pnpm install
pnpm dev:web:demo
# Open http://localhost:3001/sandbox
The sandbox includes fleet overview, trace investigation, run diff, session replay, regression detection, experiments, budgets, SLAs, prompt management, and an SDK ingestion simulator.
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