MCP Hub
Back to servers

foxhound

Validation Failed

MCP server for querying Foxhound traces from Claude Code, Cursor, and other MCP clients

Registry
Stars
2
Updated
Apr 13, 2026
Validated
Apr 14, 2026

Validation Error:

Process exited with code 1. stderr: npm error code E404 npm error 404 Not Found - GET https://registry.npmjs.org/@foxhound%2fapi-client - Not found npm error 404 npm error 404 '@foxhound/api-client@0.1.0' is not in this registry. npm error 404 npm error 404 Note that you can also install from a npm error 404 tarball, folder, http url, or git url. npm error A complete log of this run can be found in: /home/runner/.npm/_logs/2026-04-14T03_02_34_960Z-debug-0.log

Quick Install

npx -y @foxhound-ai/mcp-server

Foxhound

Compliance-grade observability for AI agent fleets.
Trace every decision. Evaluate every response. Budget every dollar.

CI License npm PyPI MCP Server

Docs · Quickstart · Integrations · Self-Hosting · Changelog


Why Foxhound?

AI agents make thousands of autonomous decisions per session — calling tools, invoking LLMs, branching workflows. When something goes wrong, you need more than logs. You need full decision traceability: what the agent decided, what data it had, and why it diverged from the expected path.

Foxhound is an open-source observability platform purpose-built for production AI agent systems. It gives your team structured tracing, automated evaluation, cost controls, and compliance tooling — all self-hostable and extensible.

Features

CategoryCapabilities
TracingStructured span trees for every agent run · Session Replay · Run Diff to compare executions side-by-side
EvaluationLLM-as-judge evaluators · Dataset curation from production traces · Experiment runner with A/B comparison
Cost & SLAPer-agent cost budgets with SDK callbacks · SLA monitoring for duration and success rate targets · Automated alerting
Regression DetectionBehavioral baselines per agent version · Automated drift detection across deployments
Prompt ManagementVersioned prompt templates · Label-based promotion (staging → production)
MCP Server37 AI debugging tools for Claude Code, Cursor, Windsurf, and other MCP clients
Quality GateGitHub Actions integration — block PRs that fail eval thresholds
AlertsMulti-channel notifications: Slack, PagerDuty, GitHub Issues, Linear, webhooks
Audit & ComplianceFull audit log · Multi-tenant data isolation · API key management (SHA-256 hashed)

Quickstart

1. Start the platform

git clone https://github.com/caleb-love/foxhound.git && cd foxhound
pnpm install
docker compose -f docker-compose.dev.yml up -d   # Postgres + Redis
cp apps/api/.env.example apps/api/.env            # set JWT_SECRET
pnpm --filter @foxhound/db db:migrate
pnpm dev                                          # API → localhost:3001

2. Instrument your agent

pip install foxhound-ai

from foxhound import FoxhoundClient

fox = FoxhoundClient(api_key="fh-...", endpoint="http://localhost:3001")

async with fox.trace(agent_id="support-agent") as tracer:
    span = tracer.start_span(name="tool:search", kind="tool_call")
    span.set_attribute("query", "refund policy")
    span.end()

3. Explore traces

Open the Foxhound dashboard to browse span trees, replay sessions, and diff runs.

Integrations

First-party SDKs for Python and TypeScript with auto-instrumentation for major agent frameworks.

FrameworkSDKAuto-instrumentation
LangGraphPython
CrewAIPython
Pydantic AIPython
OpenAI AgentsPython
Google ADKPython
Claude Agent SDKPython + TypeScript
AWS Bedrock AgentCorePython
MastraTypeScript
OpenTelemetryAnyProtocol-level

Install:

pip install foxhound-ai                              # Python
pip install "foxhound-ai[langgraph]"                 # with LangGraph auto-instrumentation
pip install "foxhound-ai[crewai,pydantic-ai]"        # multiple extras
npm install @foxhound-ai/sdk                         # TypeScript

Tooling

ToolInstallWhat it does
Python SDKpip install foxhound-aiTrace, evaluate, and budget AI agents
TypeScript SDKnpm install @foxhound-ai/sdkSame, for Node.js / Deno / Bun
CLInpm install -g @foxhound-ai/cliQuery traces, manage keys and alerts from the terminal
MCP Servernpm install -g @foxhound-ai/mcp-server37 debugging tools for AI-native IDEs
GitHub Actioncaleb-love/foxhound-quality-gateCI quality gate — block PRs that fail eval

Architecture

                  ┌──────────────────────────────────────────────┐
  SDKs / OTel ──▶ │  API (Fastify)          Worker (BullMQ)      │
                  │  80 endpoints            7 queue processors   │
                  │  ┌───────────┐           ┌────────────────┐  │
                  │  │ Traces    │           │ Evaluator      │  │
                  │  │ Evals     │           │ Experiment     │  │
                  │  │ Budgets   │           │ Cost Monitor   │  │
                  │  │ Prompts   │           │ SLA Check      │  │
                  │  │ Alerts    │           │ Regression     │  │
                  │  └───────────┘           └────────────────┘  │
                  │          │                       │            │
                  │     PostgreSQL 16           Redis / BullMQ    │
                  └──────────────────────────────────────────────┘

Stack: TypeScript · Node.js 20 · pnpm · Turborepo · Fastify · Drizzle ORM · PostgreSQL 16 · Redis · BullMQ · Stripe

Self-Hosting

Foxhound is designed to run on your infrastructure. You need PostgreSQL 16+ and Node.js 20+. Redis is required for the worker (evaluation, cost monitoring, SLA checks).

See apps/api/.env.example for all configuration options.

Dashboard: caleb-love/foxhound-web

Development

pnpm build            # Build all packages (Turborepo)
pnpm test             # Run all tests (Vitest)
pnpm test:coverage    # Tests with coverage report
pnpm lint             # Lint all packages
pnpm typecheck        # TypeScript type checking
pnpm format           # Format with Prettier

Monorepo structure

apps/api/               Fastify REST API
apps/worker/            BullMQ background workers
packages/db/            Drizzle schema + migrations
packages/sdk/           TypeScript SDK
packages/sdk-py/        Python SDK
packages/cli/           CLI
packages/mcp-server/    MCP debugging tools
packages/api-client/    Typed HTTP client (shared)
packages/billing/       Stripe + entitlements
packages/notifications/ Multi-provider alert dispatch
packages/types/         Shared type definitions
docs-site/              Docusaurus documentation (22 pages)

Security

  • JWT authentication with 30-day expiry
  • API keys SHA-256 hashed, never stored plaintext
  • HttpOnly / Secure / SameSite cookies
  • CSP headers via Helmet
  • Rate limiting on all endpoints
  • All data queries scoped by org_id — strict multi-tenant isolation
  • Parameterized queries throughout (no string interpolation)

To report a vulnerability, open a security advisory.

Contributing

We welcome contributions. See CONTRIBUTING.md for guidelines.

License

MIT — see LICENSE.

Reviews

No reviews yet

Sign in to write a review