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

117

工具排名

按实测增长排序，并在不同来源间归一化。 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. 1

    智能体
    活跃

    Generic DeepSeek Harness (dsh) plugins: A2A protocol server, session storage mirror, and Langfuse observability.

    TypeScript许可证: MIT
    • agent2agent
    • ai-agents
    • cordis
    • deepseek-harness
    • dsh

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  2. 2

    MCP
    活跃

    A typed, replayable, budget-aware programming language for LLM agents. Compiles to Python & TypeScript.

    Shell
    • a2a
    • agent-framework
    • agentic-ai
    • ai
    • ai-agents

    github实测增长打开来源 ↗

    2GitHub 星标稳定
  3. 3

    技能
    活跃

    Claude Code skills where every entry ships receipts — accuracy-gated benchmarks against baseline and placebo, rejects published

    JavaScript许可证: MIT
    • agent-skills
    • ai-agents
    • anthropic
    • benchmark
    • claude

    github实测增长打开来源 ↗

    安装 git clone https://github.com/sjh9714/skill-receipts ~/.claude/skills/skill-receipts

    2GitHub 星标稳定
  4. 4

    技能
    活跃

    Production patterns for agent orchestrators, harnesses, evals, MCP/A2A interoperability, memory, provider routing, and LLM usage metering.

    Python许可证: MIT
    • a2a
    • agent-orchestration
    • agent-skills
    • ai-agents
    • developer-tools

    github实测增长打开来源 ↗

    安装 git clone https://github.com/ssheleg/agent-stack ~/.claude/skills/agent-stack

    2GitHub 星标+1 (+100.0 %)
  5. 5

    智能体
    活跃

    A lightweight Python library that decouples agentic runtime from applications it builds

    Python许可证: MIT
    • a2a
    • agent-runtime
    • agents
    • observability
    • portability

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  6. 6

    技能
    活跃

    Multi-agent quality gate skill for Claude Code that researches, reviews, tests and challenges AI-generated work before the final answer.

    • agentic-coding
    • agentic-workflow
    • ai-agent-skills
    • ai-code-review
    • ai-workflows

    github实测增长打开来源 ↗

    安装 git clone https://github.com/ma-nucho-pro/supervisor-skill-claude ~/.claude/skills/supervisor-skill-claude

    2GitHub 星标稳定
  7. 7

    其他
    休眠

    Multi-agent customer support system with Google ADK & Gemini 2.5 Flash Lite. Kaggle capstone demonstrating 11+ concepts. Automates 80%+ queries, <10s response time.

    Python许可证: MIT
    • a2a-protocol
    • agent-evaluation
    • google-adk
    • multi-agent-system
    • observability

    github实测增长打开来源 ↗

    2GitHub 星标稳定
  8. 8

    智能体
    休眠

    A Multi-Agent System (MAS) evaluation framework using PydanticAI that generates and evaluates scientific paper reviews through a three-tiered assessment approach: traditional metrics, LLM-as-a-Judge, and graph-based complexity analysis.

    Python许可证: Apache-2.0
    • a2a-protocol
    • agent-evaluation
    • agentbeats
    • ai-agents
    • benchmarks

    github实测增长打开来源 ↗

    2GitHub 星标稳定
  9. 9

    智能体
    活跃

    Deterministic crash tests for agent control planes: paired worlds, hard stops, budgets, scopes, approvals, and hash-linked receipts.

    TypeScript许可证: Apache-2.0
    • a2a
    • agent-evaluation
    • ai-agents
    • benchmark
    • deterministic-replay

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  10. 10

    智能体
    活跃

    Agent 降智检测与自愈公评网络 — an immune system for the AI agent society

    TypeScript许可证: MIT
    • a2a
    • agent
    • agent-guard
    • ai-sre
    • drift-detection

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  11. 11

    其他
    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.

    Python
    • a2a
    • agent-runtime
    • ai-infra
    • deepseek
    • fastapi

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  12. 12

    智能体
    休眠

    The reliability and interoperability layer for AI agents. A2A-native. MCP-bridged.

    Python许可证: Apache-2.0
    • a2a
    • agent-framework
    • agents
    • ai
    • ai-safety

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  13. 13

    智能体
    活跃

    A local A2A event and continuity service for agent applications, with structured history, provenance-safe transcripts, literal search, and natural-language queries.

    Python许可证: BSD-2-Clause
    • a2a
    • agent-memory
    • ai-agents
    • event-sourcing
    • fastapi

    github实测增长打开来源 ↗

    1GitHub 星标+1
  14. 14

    技能
    活跃

    Skill instalable para Claude Code que mide si tu agente cumple las reglas establecidas. Lee el historial de sesiones de tu disco y devuelve un porcentaje por regla de cumplimiento. Sin instalar nada, sin red, biblioteca estándar.

    Python许可证: MIT
    • ai-evaluation
    • claude
    • claude-code
    • claude-code-plugin
    • claude-code-skill

    github实测增长打开来源 ↗

    安装 /plugin marketplace add jleonceo/skill-adherencia-reglas

    1GitHub 星标稳定
  15. 15

    技能
    活跃

    Mide si tu agente de IA cumple las reglas que le escribiste. Lee el historial local de Claude Code y devuelve un porcentaje por regla. Sin instalar nada, sin red, biblioteca estándar.

    Python许可证: MIT
    • agent-evaluation
    • ai-agents
    • claude-code
    • claude-code-skill
    • claude-code-skills

    github实测增长打开来源 ↗

    安装 git clone https://github.com/jleonceo/adherencia-reglas ~/.claude/skills/adherencia-reglas

    1GitHub 星标稳定
  16. 16

    智能体
    休眠

    Contract-based testing for LLM agents: hybrid evaluation, multi-agent + A2A, deterministic replay.

    Python许可证: MIT
    • a2a
    • agents
    • contracts
    • evaluation
    • llm

    github实测增长打开来源 ↗

    0GitHub 星标稳定
  17. 17

    MCP
    活跃

    8 MCP SMB products — standalone AI servers for SMBs: Guardrails, FinOps, Observability, Router, Trust Score, Memory, ThinkSecure, A2A Lite. 37 tools, TNC credits billing.

    Python许可证: MIT
    • a2a-protocol
    • ai-governance
    • finops
    • guardrails
    • mcp

    github实测增长打开来源 ↗

    1GitHub 星标稳定
  18. 18

    智能体
    活跃

    面向长程研发任务的 Java Agent Harness，支持 A2A 跨语言协作、可恢复执行、上下文工程与 Eval 驱动开发。

    Java许可证: Apache-2.0
    • a2a
    • agentic-ai
    • ai-agents
    • developer-tools
    • evals

    github实测增长打开来源 ↗

    0GitHub 星标稳定
  19. 19

    智能体
    活跃

    Local-first control plane for durable, governed agent teams: DAG orchestration, approvals, memory, signed skills, trace contracts, and realtime.

    Python许可证: Apache-2.0
    • a2a-protocol
    • agentic-ai
    • ai-agents
    • fastapi
    • llmops

    github实测增长打开来源 ↗

    0GitHub 星标稳定
  20. 20

    智能体
    活跃

    A2A green-agent orchestrator for evaluating agents on the AppWorld benchmark, built on the AgentBeats SDK

    Python
    • a2a-protocol
    • agentbeats
    • appworld
    • evaluation
    • llm-agents

    github实测增长打开来源 ↗

    0GitHub 星标稳定
  21. 21

    其他
    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

    Python许可证: MIT
    • a2a-protocol
    • agent-framework
    • ai
    • causal-memory
    • event-stream

    github实测增长打开来源 ↗

    0GitHub 星标稳定
  22. 22

    智能体
    休眠

    A Python proof-of-concept for tracing multi-turn Agent-to-Agent (A2A) conversations as a single unified MLflow trace for LLM observability and evaluation.

    Python
    • a2a
    • agent-to-agent
    • ai-agents
    • conversational-ai
    • llm-evaluation

    github实测增长打开来源 ↗

    0GitHub 星标稳定
  23. 23

    MCP
    活跃

    Open-source, self-hosted control plane for OpenClaw and Hermes AI-agent fleets on Docker/Kubernetes — REST, CLI, and MCP.

    TypeScript许可证: Apache-2.0
    • agent-management
    • agent-ops
    • agent-orchestration
    • ai-agent-platform
    • ai-agents

    github实测增长打开来源 ↗

    50GitHub 星标+2 (+4.2 %)
  24. 24

    技能
    活跃

    Real-time execution trace and cost intelligence for Claude Code

    TypeScript许可证: MIT
    • agentic-workflow
    • ai-productivity
    • anthropic
    • anthropic-claude
    • claude

    github实测增长打开来源 ↗

    安装 /plugin marketplace add DeibyGS/claudestat

    34GitHub 星标+1 (+3.0 %)
  25. 25

    技能
    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.

    Shell许可证: MIT
    • anthropic
    • claude
    • claude-code
    • claude-code-plugin
    • evals

    github实测增长打开来源 ↗

    安装 /plugin marketplace add rennf93/opus-fable-playbook

    33GitHub 星标稳定