AIエージェント
smolvm
An embeddable, portable, branchable microVM that runs AI agents safely on your own machine.
★ 6.6k スターApache-2.0
ガイドを見る →julepは、julep-ai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: julep-ai/julep）。GitHubスターは6,577、フォークは968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8月6日です。
Durable, composable AI agents — flows that crash and resume, retry safely, and explain every step.
Sketch the multi-step agent flow you need — the tools it may call, the decisions it must make, and where it must be able to resume after a failure.
Install the julep Python package (the current release line is pre-release, so use the documented --pre flag).
Register your tools or MCP servers and keep production secrets in your deployment environment or Kubernetes Secrets, as the docs require.
Define a @flow, dry-run it locally with in-memory tools, then deploy it and watch each step execute.
Flows are built to crash and resume and to retry safely, so long-running agent work is not lost when a process dies.
Define flows in Python: registered tools and pure functions, think/cond/switch/each/reschedule steps, and record merging with the | operator.
Every step of a flow is recorded so you can see what the agent did and why, rather than getting an opaque final answer.
Call MCP servers from inside a flow with mcp_tool(server, tool), alongside natively registered tools.
The julep CLI supports plan, apply and status with composable selectors (tags, state, graph traversal) across environments.
dry_run executes a flow locally with in-memory tools and deterministic fake reasoners, so you can test logic before deploying.
Clone Julep to explore the source, examples and docs.
git clone https://github.com/julep-ai/julep.git
cd julepInstall the current Julep Python package from PyPI.
pip install --pre julepRun one of the bundled @flow examples locally to see a durable flow execute step by step.
python examples/episode_summary_flow.py # see README for the exact commandjulep-ai maintains Julep; documentation is published at docs.julep.ai.
The open-source project is licensed under Apache-2.0. Julep also offers a hosted platform, but the repository code itself is free to use under that license.
Flows call LLMs through reasoners, so in production you typically supply your own model provider credentials; local dry runs can use deterministic fake reasoners instead.
Both build graph-like agent workflows. Julep emphasizes durable execution — crash-and-resume flows, safe retries and step-by-step explanations — managed through its CLI and deployment model.
pip install --pre julep, then work through the Quickstart and the example flows in the examples directory.
An embeddable, portable, branchable microVM that runs AI agents safely on your own machine.
A control plane with a durable state kernel that keeps long-horizon agents moving across sessions.
Vercel’s SDK for durable, reliable, observable apps and AI agents in TypeScript.
All your agents and workspaces in one place — on every device you own.
A secure, fast sandbox runtime where AI agents can execute safely.
Give your coding agents a persistent memory that survives every session.
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 — julep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julep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。
An embeddable, portable, branchable microVM that runs AI agents safely on your own machine.
Open-source framework for building agentic apps in JavaScript, Go, Dart, and Python, built and used in production by Google
A payments protocol for the internet — built on plain HTTP.
An MCP server and CLI that lets AI agents build, test and run iOS and macOS projects.
AppAgent — Tencent's multimodal agent that learns smartphone apps by exploration, then operates them.
Policy, identity, and sandboxing guardrails for AI agents.