GitHub
nanoAgent
观察池 · 暂无正式排名
该项目当前未满足“两个有效维度 + 两种数据源”的主榜门槛。
— 未排名
可观测采用度缺失 —
动量当前有效 · 2026-09-12 6
关注度当前有效 · 2026-09-12 0
信号可信度 依据当前有数据的独立评分维度数量计算。
中2/3 · 1 种数据源
项目介绍
"The question is not what you look at, but what you see." — Henry David Thoreau
The simplest way to build an agent that can interact with your system.
A minimal implementation of an AI agent using OpenAI's function calling. The agent can execute bash commands, read files, and write files.
If you want to learn more (e.g. what MCP is, and how to fetch tools in a more modern way), see:
The core is just a loop: call model → execute tools → repeat.
Recent hardening keeps the loop running even when a tool call contains malformed JSON arguments or references an unknown tool; those cases are returned to the model as explicit tool errors instead of crashing the agent.
⏺ Like a single seed that…
各数据源
744 Star
- Star 744
- Fork 285
- 提交 11
- 发布 0