AIエージェント
SWE-bench
The benchmark that grades AI on real GitHub issues.
★ 6.0k スターMIT
ガイドを見る →tau2-benchは、sierra-research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: sierra-research/tau2-bench）。GitHubスターは2,186、フォークは562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7日です。
A benchmark for tool-agent-user interaction in real-world domains.
Start with one concrete task rather than a grand architecture.
Clone the repo and check the examples folder for a quickstart.
Install packages, set environment variables, and point it at your model.
Execute the demo workflow, then adapt it to your own data.
Tool, agent, and user interaction.
Realistic task settings.
Dialogue agents under test.
From Sierra Research.
Inspectable benchmark design.
Free to use and extend.
Download the source and move into the project folder.
git clone https://github.com/sierra-research/tau2-bench.git
cd tau2-benchCreate a virtual environment and install the requirements.
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txtTreat these as starting points - the README documents the exact commands.
python examples/quickstart.pysierra-research/tau2-bench is its home on GitHub, with 562 forks so far. The contributors tab and recent commits show how actively it is maintained.
Yes - it is MIT licensed, so you can use, modify, and distribute it freely, including in commercial projects.
Running the evaluation is free; the model under test may use your own paid API keys.
tau2-bench evaluates tool-agent-user interaction in conversational domains; ClawBench targets browser agents on daily web tasks. Use both if your agent talks and browses.
Follow the README quickstart, then browse the examples directory for patterns that match what you want to build.
The benchmark that grades AI on real GitHub issues.
Benchmark computer-use agents on long, real-world desktop tasks.
Pinvou Agent — open-source desktop AI agent for tools, files, knowledge, workflows and real deliverables
An auto-company that works 24/7 on your own PC — 14 agent roles, a dashboard, real workflows.
Infisical's credential vault and HTTP proxy, so AI agents call APIs without ever holding your real secrets.
WebArena-Verified: the audited, deterministically scored release of the WebArena web-agent benchmark.
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 — tau2-bench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tau2-bench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。
Open Cowork: a desktop AI agent app with one-click Claude Code, MCP tools, skills, and VM sandboxing.
Shrimp Task Manager: an MCP server that turns natural language into structured, dependent dev tasks.
Self-hosted AI vulnerability research: agents find, validate, deduplicate, and rank real security issues.
One open platform to simulate, evaluate, guardrail, trace, and optimise LLM and agent apps, self-hostable.
Unla MCP Gateway: turn existing APIs and MCP servers into MCP endpoints with zero code changes.
Open-source AI assistant framework with skills and an OpenClaw-style agent core.