Read-Only Kernel Telemetry as MCP Tools: A Design NoteDesign constraints (left) and the seven existing tool calls (right) for a read-only MCP server over an eBPF kernel trace database. TL;DR An MCP server over kernel telemetry has a specific shape: read-Jul 13 · 6 min readingero.hashnode.dev
Top of Stack, Bottom of Stack: Where RUM Ends and eBPF BeginsThe five layers of the production observability stack. RUM observes user-visible events at the top. eBPF observes kernel events at the bottom. The middle three need different tools. TL;DR The productiJul 6 · 6 min readingero.hashnode.dev
The Seventh Agent: Counting Privileged Processes on a Real GPU HostA real audit of host-level agents already running on a typical GPU host. Adding kernel-level GPU observability is a decision about whether to add a seventh privileged process or to put the instrumentaJun 29 · 7 min readingero.hashnode.dev