AIエージェント
helicone
Open-source LLM observability: monitor, evaluate, and experiment.
★ 6.2k スターApache-2.0
ガイドを見る →opikは、comet-mlが公開しているオープンソースのAIエージェントプロジェクトです（GitHub: comet-ml/opik）。GitHubスターは22,451、フォークは1,849で、Apache-2.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月8日です。
Open-source LLM observability: trace, evaluate and monitor apps and agents
Use Comet's cloud for the fastest start, or self-host the full platform with Docker Compose or Helm.
Install the Python SDK with pip, or clone the repository to self-host the server.
Run opik configure, add the tracing decorator to your functions, and point the SDK at your workspace.
Run your app once and open the Opik dashboard to inspect the full trace and score it.
Logs LLM calls including nested function and agent traces, so you can replay exactly what happened.
Score outputs with automated evaluation methods and run experiments against datasets.
Monitor quality, usage and feedback from live traffic with production-ready dashboards.
Built for agentic workflows and RAG systems, not just single completions.
Works across ecosystems - the repo highlights LangChain and LlamaIndex among its integrations - without locking you into one framework.
Deploy locally with Docker Compose or at scale with Kubernetes and Helm, or use the Comet cloud.
Get the source code from GitHub and enter the project directory.
git clone https://github.com/comet-ml/opik.git
cd opikAdd Opik to your Python project and connect it to your workspace or local server.
pip install opik
opik configureClone the repo and start the full Opik suite locally with the provided script (run inside the clone).
./opik.shIt is maintained by Comet (the company behind Comet ML) as an open-source project on GitHub.
Yes - Opik is Apache-2.0 licensed, and the README stresses the full platform can be self-hosted free without a commercial license. Comet also offers a hosted cloud option.
Opik itself needs no paid key when self-hosted; the LLM calls you trace still use your own provider keys, and the cloud option needs a Comet account.
Compared with Langfuse, Opik positions itself as a fully Apache-2.0, self-hostable platform with evaluation and optimization built in; the README includes a comparison table against alternatives.
Run pip install opik and opik configure, decorate a function with the track decorator, and view the trace in the dashboard - or self-host with ./opik.sh after cloning.
Open-source LLM observability: monitor, evaluate, and experiment.
Open-source observability for AI agents: trace, evaluate, guard, and improve.
One open platform to simulate, evaluate, guardrail, trace, and optimise LLM and agent apps, self-hostable.
OpenTelemetry-native observability for LLM apps.
Observability and evaluation for AI apps — see what your agents do.
Code, build and evaluate LLM agents with full MCP, Skills, ACP and A2A support.
LibreAIカタログに掲載されたAIエージェントプロジェクトの中で最もスターの多いもの — opikが合わなかった場合に検討したい、現実的な候補リストです。
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
カタログの中でopikとスター数が近いAIエージェントプロジェクト — 前後の棚もそのまま見て回れます。
NVIDIA's reference stack for running AI agents more securely inside OpenShell sandboxes
Yohei Nakajima's autonomous task agent - now a self-building agent framework in Python
An MCP server that lets Claude, Cursor and Windsurf build n8n workflows for you
Stanford's Generative Agents simulation - 25 AI characters living in a virtual town
A framework for building agents and services
Swarm — OpenAI's experimental framework for lightweight multi-agent orchestration with agents and handoffs, MIT