工具说明为英文。
LLM 可观测性
LLM 应用的监控、追踪、评估和质量管理。
用途
活跃度
排序方式
123
工具排名
Ranked by creation date, newest first; undated entries come last. 原始值保留各自的数据窗口；计算变化需要 7 天内至少两次测量。
- 1活跃
Architecture-first Python scaffold for an auditable multi-agent corporate credit desk using A2A, MCP, deterministic credit policies, model routing, and OpenTelemetry.
github实测增长打开来源 ↗
0GitHub 星标稳定 - 2活跃
AI Guardian
MCPGoverned local-LLM observability: model policy, prompt scanner, capture proxy, 21 tools.
mcp实测增长打开来源 ↗
安装
claude mcp add ai-guardian -- uvx ai-guardian-aiops0GitHub 星标稳定 - 3782GitHub 星标+1 (+0.13 %)
- 4活跃
🔑 HomeStream · 家园·流 — 零成本自托管多Agent协作框架，通往AI世界的那把钥匙 | Zero-cost self-hosted multi-agent framework — The key to AI world
github实测增长打开来源 ↗
0GitHub 星标稳定 - 5活跃
bakeoff
技能Turn one decision into a judged tournament of solutions, then pick the best — a Claude Code skill that generates candidates, auto-derives the rubric, judges independently, and returns a defensible winner.
github实测增长打开来源 ↗
安装
/plugin marketplace add CoriChui/bakeoff10GitHub 星标稳定 - 6活跃
The Fable Workflow: how Claude Fable 5 worked, distilled into skills any model can run, with the eval that keeps it honest. Think / act / prove.
github实测增长打开来源 ↗
安装
git clone https://github.com/Sahir619/fable-method ~/.claude/skills/fable-method2 273GitHub 星标+11 (+0.49 %) - 7活跃
Self-hosted AI SRE for Kubernetes — zero-instrumentation eBPF observability plus a copilot that fixes issues through guardrailed, self-verifying actions. BYO-LLM, air-gapped capable.
github实测增长打开来源 ↗
131GitHub 星标-4 (-3.0 %) - 8活跃
Open benchmark for Claude Code SEO skills — real headless execution against fixture sites with planted-defect answer keys. Deterministic scoring, pre-registered rubric.
github实测增长打开来源 ↗
安装
git clone https://github.com/aleclindz/seo-skill-bench ~/.claude/skills/seo-skill-bench51GitHub 星标-5 (-8.9 %) - 9活跃
Make Claude Opus 4.8 behave like Claude Fable 5 — doctrine output style, drift-catching hooks, and an eval loop against golden Fable transcripts. Claude Code plugin.
github实测增长打开来源 ↗
安装
/plugin marketplace add rennf93/opus-fable-playbook34GitHub 星标+1 (+3.0 %) - 10活跃
Claude Code plugin marketplace — agentic-engineering (spec-driven shape→decide→execute→measure→eval with adversarial review) + github-keeper (audit/elevate READMEs and make a repo open-source-ready).
github实测增长打开来源 ↗
安装
/plugin marketplace add GiustoPiedimonte/agentic-engineering-marketplace13GitHub 星标稳定 - 11活跃
untell
其他AI-detector auditing toolkit: measures how often a detector flags genuine human writing, how stable its verdict is across seeds, and whether that verdict survives meaning-preserving edits. Detector-in-the-loop measurement harness. Claude…
github实测增长打开来源 ↗
18GitHub 星标稳定 - 12活跃
Multi-tier framework for evaluating AI agent skills with quality gates, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation that measures how skills affect agent behavior.
github实测增长打开来源 ↗
安装
git clone https://github.com/NVIDIA/SkillEvaluator ~/.claude/skills/SkillEvaluator399GitHub 星标+51 (+14.7 %) - 131GitHub 星标稳定
- 14活跃
databuff
智能体DataBuff is an AI-native APM built on Opentelemetry，with multi-agent troubleshooting out of the box.
github实测增长打开来源 ↗
649GitHub 星标+39 (+6.4 %) - 15活跃
historian
智能体A local A2A event and continuity service for agent applications, with structured history, provenance-safe transcripts, literal search, and natural-language queries.
github实测增长打开来源 ↗
1GitHub 星标稳定 - 16活跃
Terminal-first validation harness for reviewing PRDs, RFCs, strategy docs, and experiment plans.
github实测增长打开来源 ↗
1GitHub 星标稳定 - 17活跃
cap-evolve
MCPOptimize any AI agent’s skills, tools/MCP, and prompts against your own evals.
github估算动量打开来源 ↗
47GitHub 星标— - 18活跃
pyxen
智能体A lightweight Python library that decouples agentic runtime from applications it builds
github实测增长打开来源 ↗
1GitHub 星标稳定 - 19活跃
Agent Skill evaluation harness for paired variants, trace artifacts, and runner adapters
github实测增长打开来源 ↗
安装
git clone https://github.com/adewale/skill-eval-harness ~/.claude/skills/skill-eval-harness73GitHub 星标+4 (+5.8 %) - 20活跃
tulip-agents
智能体The agent framework where the model never holds the trigger — every consequential action clears your policy first, waits for a human when it matters, and lands on a record you can verify. Build on it, or put it around the agent you already…
github实测增长打开来源 ↗
1GitHub 星标稳定 - 21活跃
axiom
技能Axiom is a curated marketplace of shared plugins for Claude Code and Codex.
github实测增长打开来源 ↗
安装
git clone https://github.com/netopsengineer/axiom ~/.claude/skills/axiom5GitHub 星标稳定 - 22活跃
Local-first Agentic AI Infrastructure Platform with LLM Gateway, Agent DAG Runtime, MCP Tool Hub, A2A Agent Mesh, Local RAG, Tool Sandbox and Observability.
github实测增长打开来源 ↗
1GitHub 星标稳定
学习与参考资源
Ranked by creation date, newest first; undated entries come last. 这些资源可单独访问，不参与主要排名。
- 1活跃打开来源 ↗
50+ curated LLM observability tools PLUS 26 Agent Skills (several with runnable, unit-tested scripts) to build, evaluate, debug, secure & monitor reliable LLM apps. Tracing, evals, guardrails, LLMOps.
github安装
资源实测增长git clone https://github.com/ContextJet-ai/awesome-llm-observability ~/.claude/skills/awesome-llm-observability33GitHub 星标+3 (+10.0 %) - 2活跃打开来源 ↗
Curated + scored map of official MCP servers and agents for DevOps, Cloud, SRE, and Platform Engineering — every entry rated on production access, approval gates, and audit evidence.
github资源实测增长77GitHub 星标稳定 - 3活跃打开来源 ↗
tunelab
技能Claude Code plugin for LLM fine-tuning, distillation, and evaluation — decide whether you need fine-tuning at all, distill your LLM logs into small local models (MLX/LoRA), evaluate with held-out discipline, and learn the why at every step.
github安装
资源实测增长/plugin marketplace add rchaz/tunelab6GitHub 星标稳定