mairp

Agent Observability Stack — AI skill for Claude Code

AI community

Self-hostable observability for LLM agents + a Linux host (Intel iGPU/NPU + NVIDIA RTX 3090 eGPU aware): Prometheus + Grafana + Tempo + OTel, OpenClaw/LiteLLM/Claude Code/RAG metrics, dynamic onboardi.

How to install Agent Observability Stack

This entry records only its repository, not the path inside it, so there is no exact command to give. Open mairp/agent-observability-stack and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Observability Stack does

Alternatives in AI

Multica — Make humans and AI agents work as one team — open-source and self-hostable 47.8k ★
Ccstatusline — by sirmalloc - A highly customizable status line formatter for Claude Code CLI that displays model info, git b 5.5k ★
Cc Mirror — Create multiple isolated Claude Code variants with custom providers (Z.ai, MiniMax, OpenRouter, LiteLLM) 2.1k ★

README

agent-observability-stack

A self-hostable observability stack for **LLM agents + a Linux host (Intel iGPU/NPU + NVIDIA GPU aware)**, built as a single LAN-accessible pane of glass. Two pipelines, one Grafana:

Infra metrics (pull): CPU, RAM, disk, network + latency, containers, and accelerator telemetry — Intel iGPU/NPU and NVIDIA GPU (nvidia-smi: util, VRAM incl. per-process, temp, power, clocks) → Prometheus.
Agent telemetry (push): OpenClaw diagnostics → OTLP → OpenTelemetry Collector → Grafana Tempo (traces) + Prometheus (metrics); LLM tokens, cost, latency, cache from a LiteLLM proxy; and local llama.cpp inference (/metrics: decode/prefill tok/s, queue depth, and an MTP speculative-decoding acceptance signal).

New agents, containers, and accelerators **onboard themselves** — no per-target config edits.

Demo

Full walkthrough (~72s) — the dashboards end to end: infra (host, accelerators, containers), the agents pipeline (cost, tokens, latency, call outcomes, traces, live agent inventory), the **Claude Code & RAG** dashboard, and the **QMD Vector Space** (RAG memory corpus + benchmark queries projected to 2D/3D).

https://github.com/user-attachments/assets/ef08dcad-d20c-43eb-aa14-42b8cd2c014c

Architecture

INFRA METRICS (pull)                       AGENT / AI TELEMETRY (push)
 node_exporter ────┐                        OpenClaw gateway          Claude Code (OTel)
 cadvisor         ─┤                          ├─ diag-prometheus ─(loopback textfile)─┐   │
 blackbox         ─┤                          │                                       │   │ OTLP
 intel igpu/npu   ─┼─► Prometheus ◄───────────┤                                       │   ▼
 nvidia gpu       ─┤        ▲                  └─ diag-otel ──OTLP─┐                   └► OTel Collector
 llama.cpp /metrics┤        │                                     ▼                        │
 litellm /metrics