inspektor-gadget/ig-mcp-server

Overview

A Go-based MCP server that exposes Inspektor Gadget's container and Kubernetes debugging capabilities through an AI interface. It allows developers to inspect workloads, trace system calls, and analyze network activity using natural language queries.

Best for

Best for
Developers debugging container and Kubernetes workloads who want a conversational interface to eBPF-based introspection tools.

Use cases

• Debug container crashes by querying recent syscall activity
• Trace network connections from a specific pod without kubectl exec
• Analyze file system operations of a misbehaving Kubernetes workload

How to use

Tools exposed

• ig_deploy
• ig_gadgets
• gadget_trace_dns
• gadget_trace_tcp
• gadget_trace_exec
• gadget_trace_open
• gadget_trace_signal
• gadget_trace_bind
• gadget_snapshot_process
• gadget_snapshot_socket
• gadget_top_file
• gadget_top_tcp
• gadget_top_blockio
• gadget_profile_blockio
• gadget_profile_tcprtt
• gadget_trace_capabilities
• gadget_advise_seccomp
• gadget_audit_seccomp
• gadget_trace_lsm
• gadget_traceloop

Tested with

VS Code, GitHub Copilot CLI, Claude Code

Notes

A Go-based MCP server that exposes Inspektor Gadget’s container and Kubernetes debugging capabilities through an AI interface. It allows developers to inspect workloads, trace system calls, and analyze network activity using natural language queries.

25 stars on GitHub. Last updated 2026-05-22. Licensed Apache-2.0.

Use cases

• Debug container crashes by querying recent syscall activity
• Trace network connections from a specific pod without kubectl exec
• Analyze file system operations of a misbehaving Kubernetes workload

Pros

• Leverages Inspektor Gadget’s powerful eBPF-based introspection without manual CLI commands
• Integrates with AI assistants for interactive debugging sessions
• Open source and written in Go, easy to extend or audit

Cons

• Requires Inspektor Gadget to be installed and configured on the cluster
• Small community (25 stars) means limited documentation and support
• AI interface adds latency compared to direct CLI usage

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.