开发者工具DEVELOPER TOOLS· LEARNING
ai-boost/awesome-harness-engineering
收集智能体开发工具与模式的清单,方便开发者查阅。A curated list of tools and patterns for building, evaluating, and controlling every layer of an AI agent harness.
根据公开文档整理,未运行实测;费用以项目官网为准。Based on public documentation, without runtime testing. Check the project site for pricing.
能力CAPABILITIES
Find tools for memory, permissions, and evaluation in one list
Compare patterns for context delivery and task decomposition
Set up observability to see what your helpers does
Control tool access and browser interactions
Build helpers loops with planning and compaction
上手GET STARTED
文档里没有找到可识别的安装命令。这类项目按发布规则不会进入榜单,只能在这里查看。No recognisable install command was found in the docs. Projects like this never enter the ranked list under our publishing rules — they are viewable here only.
文档未提供安装说明No install commands are provided in the evidence; the list is meant to be browsed and used as a reference.
使用限制LIMITATIONS
以下要求与风险来自文档静态检查,未执行项目代码;扫描没有命中,不代表使用安全。Risk here is evidence, not an alarm. These findings come from static scanning of the docs; we never execute a project’s code.
限制LIMITS
需检查Needs review
静态扫描STATIC SCAN
01控制浏览器Controls a browserreadme
- [context-mode](https://github.com/mksglu/context-mode) — MCP server that intercepts raw tool output before it enters the context window, s
静态扫描未发现高风险项,但有一项 medium 风险:文档中提到的 context-mode 这个 MCP 服务会在工具输出进入上下文窗口前拦截原始输出,涉及控制浏览器的行为,使用前需要留意。A static scan found a medium-risk item: one listed tool can control a browser. Review that tool's permissions before using it in production.
项目动态PROJECT ACTIVITY
首次发现FIRST SEEN
03 OCT 202603 OCT 2026
增长GROWTH
+10.5%+10.5%
4.7k → 5.2k stars · 45 个快照4.7k → 5.2k stars · 45 snapshots
状态STATUS
持续活跃Active
质量 C+Quality C+
项目创建不到一年(2026年3月)已获得4758星,社区关注度高4758 stars and 633 forks show strong community interest
热度分MOMENTUM
自 2026-10-08 起追踪 25 天。热度分从 38.4 到 44,净变化 +5.6。峰值 55.3,出现在 2026-10-14。Tracked 25 days since 2026-10-08. Momentum went from 38.4 to 44, a net change of +5.6. The peak was 55.3 on 2026-10-14.
折线从左到右按追踪日期排列,表示热度分。The line reads left to right by tracked date and shows momentum.
数据说明:走势来自 Signal Garden 每日追踪的热度分;该项目尚未出现在公开每日榜单上。Data note: the trend uses Signal Garden’s daily tracked momentum scores; only dates marked “Ranked” appeared on the public daily list.
共 25 天追踪记录,这里列出最近 24 天。25 tracked days; the latest 24 days are listed.
同类RELATED
下面这些项目与它有相同的语言、分类或仓库主题。热度分是各自最近一次上榜时的分数。These projects share a language, category, or repository topic. Momentum is the score from each project's latest ranking appearance.
相同主题 ai-agents、context-engineering · 热度分 59.8 · 2026-10-26shared topic ai-agents, context-engineering · momentum 59.8 · 2026-10-26
相同主题 agent-memory、ai-agents · 热度分 70.5 · 2026-10-22shared topic agent-memory, ai-agents · momentum 70.5 · 2026-10-22
相同主题 ai-agents · 热度分 15 · 2026-10-04shared topic ai-agents · momentum 15 · 2026-10-04
相同主题 ai-agents、context-engineering · 热度分 55.7 · 2026-10-13shared topic ai-agents, context-engineering · momentum 55.7 · 2026-10-13
同为 Python · 热度分 57.7 · 2026-10-27same language, Python · momentum 57.7 · 2026-10-27
相同主题 ai-agents、mcp · 热度分 81.6 · 2026-10-21shared topic ai-agents, mcp · momentum 81.6 · 2026-10-21
读者WHO IT'S FOR
用 Python 做 agent-harness、agent-memory、agent-orchestration 的团队会用它处理这件事:Awesome Harness Engineering Curated resources, patterns, and templates for building reliable AI agent harnesses。需要留在 Python 而不是站内同类 xerj 所用的 Rust 时优先选它,热度走势为峰值回落。Teams using Python on agent-harness, agent-memory, agent-orchestration use it for this problem: Awesome Harness Engineering Curated resources, patterns, and templates for building reliable AI agent harnesses. Prefer it over the in-site peer xerj when the stack stays Python rather than Rust, the momentum shape is peak-then-fade.
{
"agents": [
{
"agent": "claude-code",
"evidence": "readme: …eering\">中文</a> </p> </div> Claude Code skills for many of the proble…"
},
{
"agent": "codex",
"evidence": "readme: …first world. - [Unrolling the Codex Agent Loop](https://openai.com/inde…"
},
{
"agent": "cline",
"evidence": "readme: …laude Code, Codex CLI, Aider, Cline, OpenHands, and SWE-agent, it…"
},
{
"agent": "gemini-cli",
"evidence": "readme: …sses (Claude Code, Codex CLI, Gemini CLI, Mistral Vibe, OpenHands, Aid…"
},
{
"agent": "vscode",
"evidence": "readme: …ness Behind GitHub Copilot in VS Code](https://code.visualstudio.co…"
},
{
"agent": "opencode",
"evidence": "readme: …ed across Claude Code, Codex, OpenCode, and Cursor, and a map of sel…"
},
{
"agent": "copilot",
"evidence": "readme: …- [The Coding Harness Behind GitHub Copilot in VS Code](https://code.visu…"
}
],
"apiKey": "providing pre-built OAuth and API key authentication for 700+ APIs across 30",
"docker": null,
"taxonomy": {
"scores": {
"design": 0,
"ai-tools": 3,
"dev-tools": 1,
"indie-web": 0,
"skill-agent": 6,
"productivity": 3
},
"primary": "dev-tools",
"secondary": "learning"
},
"localRuntime": "nthropic, OpenAI, Gemini, and Ollama with native MCP, built-in observability",
"skillMdTotal": 0,
"skillMdErrors": [],
"treeTruncated": false,
"categoryScores": {
"data": 0,
"agent": 3,
"media": 0,
"design": 1,
"browser": 0,
"devtool": 1,
"security": 1,
"marketing": 0,
"productivity": 1
},
"classification": [
{
"signal": "MCP topic + README",
"weight": 0.6
},
{
"signal": "description mentions AI",
"weight": 0.2
}
],
"scannedSources": [
"readme"
],
"skillMdFetched": 0
}