Foxhound
Query Foxhound execution traces from your coding agent when debugging agent workflows or reviewing what the model actually did.
Overview
Foxhound MCP is an Operate-phase MCP server that lets agents query Foxhound traces for debugging and monitoring agent sessions.
What is this MCP server?
- stdio MCP server @foxhound-ai/mcp-server (v0.3.0) for Claude Code, Cursor, and MCP clients
- Query Foxhound traces via API key from foxhound.dev
- Configurable FOXHOUND_ENDPOINT (default http://localhost:3001) for self-hosted or remote backends
- Focused on agent run introspection, not generic application APM
- Repository subfolder packages/mcp-server on GitHub caleb-love/foxhound
- MCP server package version 0.3.0 (@foxhound-ai/mcp-server)
- Required secret FOXHOUND_API_KEY
- Default FOXHOUND_ENDPOINT http://localhost:3001
Community signal: 2 GitHub stars.
What problem does it solve?
When agent sessions go wrong, you lack a quick in-editor path to pull structured traces from Foxhound.
Who is it for?
Builders running Foxhound-instrumented Claude Code or Cursor workflows who want trace lookup without leaving the IDE.
Skip if: Teams with no Foxhound deployment who need general-purpose log search or non-agent application monitoring only.
What do I get? / Deliverables
You can ask your MCP client to fetch Foxhound trace data so root-cause analysis stays inside your coding agent workflow.
- MCP-accessible Foxhound trace queries from coding agents
- Configurable endpoint for local or hosted Foxhound backends
- In-session visibility into prior agent execution paths
Recommended MCP Servers
Journey fit
Trace retrieval supports running and improving agent systems in production-like loops, which maps to the Operate phase rather than greenfield build. Foxhound traces are observability data—monitoring and diagnosing agent behavior after sessions run.
How it compares
Foxhound trace MCP bridge, not a generic OpenTelemetry dashboard skill.
Common Questions / FAQ
Who is Foxhound MCP for?
Developers using Foxhound to record agent activity who want MCP-native trace queries from Claude Code or Cursor.
When should I use Foxhound MCP?
Use it in operate/monitor workflows when reviewing agent failures, auditing tool usage, or comparing runs after shipping agent features.
How do I add Foxhound MCP to my agent?
Install @foxhound-ai/mcp-server, set FOXHOUND_API_KEY (and optional FOXHOUND_ENDPOINT), add the stdio server to your MCP config, then query traces from the agent.