AIエージェント
cc-safety-net
Pre-execution guard that stops AI agents' destructive commands cold.
★ 1.6k スターMIT
ガイドを見る →RAMPARTは、microsoft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: microsoft/RAMPART）。GitHubスターは419、フォークは58で、MIT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6日です。
Safety testing for agents, right in pytest.
Choose one real workload to test the project against.
Clone it locally and read through the getting-started guide.
Install the stack it needs and set any environment variables.
Execute the starter example and check results before going further.
Safety tests written as pytest tests.
Built for AI agent applications.
Covers both dimensions of agent risk.
From Microsoft's AI safety efforts.
Fits existing Python test pipelines.
Permissive licensing.
Download the source and move into the project folder.
git clone https://github.com/microsoft/RAMPART.git
cd RAMPARTSet up a virtual environment and install the package.
python -m venv .venv && source .venv/bin/activate
pip install -e .Add your first test per the docs and run pytest.
pytest
# see docs for writing safety testsIt is maintained by Microsoft at microsoft/RAMPART, with 58 forks.
Yes. It is released under the MIT license, which permits commercial use, modification, and redistribution. Read the LICENSE file for the exact terms.
The framework is free; the tests themselves need model access.
RAMPART is pytest-native CI testing; garak scans models and AgentDojo evaluates attacks — different angles on agent safety.
Install the package and follow the docs to write your first safety test.
Pre-execution guard that stops AI agents' destructive commands cold.
The elegant testing framework for PHP developers and AI agents.
Autonomous AI agents for penetration testing — self-hosted, multi-agent, with a full web UI
A benchmark testing whether agents can solve long-horizon coding tasks.
Build, deploy, and monetize autonomous agents across Web3.
Your self-hostable AI second brain for answers, agents, and automations.
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 — RAMPART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RAMPART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。
The tougher, more robust successor to the MMLU benchmark.
Agentic memory for LLM agents, from a published research project.
A harness-oriented framework for production-grade agent applications.
Benchmark computer-use agents on long, real-world desktop tasks.
Snowflake's Agent World Model: synthesise 1,000 executable MCP tool-use environments for agentic RL.
WorkArena: ServiceNow's benchmark for how well web agents handle everyday knowledge-work tasks.