可自行运行的开源 AI 应用，已测试

将文档网站、GitHub 仓库和 PDF 转换为 Claude AI 技能，并自动检测冲突

测试发现：skill-seekers 3.10.0.dev0 installed and all 3857 unit/adaptor/scraper tests pass without external API keys or network dependencies. 6 分钟。

美观且高度可定制的 Claude Code CLI 状态栏，支持 powerline、主题等。

测试发现：Dependencies installed, project builds to dist/ccstatusline.js, all 1897 tests pass, and piped JSON input renders a formatted status line with model name and git info. 6 分钟。

一个自动发现热点并自行撰写日报的网站框架。将信源和精选标准替换为你自己的，即可成为你的行业热点站。

测试发现：The project installs cleanly with Node.js 24.11 and PostgreSQL 16, typechecks across 6 packages, passes all 128 backend tests and 11 frontend tests, the API server returns 200 on /openapi-v1.json, and the web server serves /, /all, /about, and /feed.xml at HTTP 200. 13 分钟。

全球首个开源企业级世界模型。

测试发现：Utopia server v0.1.0 on :1516, PostgreSQL 16 with pgvector, 21 migrations applied, 268/268 tests passing, web frontend built to web/dist. 26 分钟。

面向 AI agent 和应用的开源浏览器 API。Steel Browser 是开箱即用的浏览器沙箱，让你能够自动化操作网页，无需担心基础设施。

测试发现：With a user-local Node 22 and Chrome for Testing plus DISABLE_CHROME_SANDBOX=true, the dev-mode API runs on port 3000, /v1/health returns 200 with live Chrome, and sessions/scrape/screenshot/pdf plus external CDP client connections all worked with real results. 11 分钟。

自托管 AI 网关：多渠道多凭据统一接入，含 API keys 与订阅账号、调度、容错、请求日志和用量。

测试发现：GPT-Load 2.0 binary builds, web UI compiles, server starts on 127.0.0.1:3001, health endpoint returns HTTP 200, auto-generates AUTH_KEY, and shuts down gracefully with all database migrations completed and catalog synced. 8 分钟。

TSP 自托管、零运维的 A 股「自建策略 + 监控 + 回测」量化工作台 | LLM 能力驱使策略定制+个股分析+复盘 | 自由接入第三方数据源与个性化扩展数据 | 个人开源

测试发现：Backend server starts and serves /health (200), /api/capabilities, and /api/auth/status; all 1511 backend tests pass; frontend build succeeds via pnpm build. 24 分钟。

Grok Bot 的开源替代方案。选择自己的模型和沙盒。

测试发现：Node 22.23.2 + pnpm 9.15.0 installed; all 1853 npm packages resolved and linked; Prisma client generated; full TypeScript/Vite/Electron/Astro build completes cleanly; all 282 unit-test files pass (2605 tests); Electron 43.4.0 launches on the virtual display and stays up. 9 分钟。

基于 FunASR 的视频转录、字幕生成和 LLM 辅助剪辑工具，带有本地 Gradio UI。

测试发现：FunClip v2.2.1 is fully installed in a Python 3.12 venv with all dependencies, all 90 tests pass, the CLI recognition+clipping pipeline processes a real video with Paraformer-Large ASR and produces valid clipped MP4 output, and the Gradio web interface launches and serves HTTP 200 on localhost. 49 分钟。

AI 原生的 JSON Gantt 图表引擎。无依赖的交互式时间线，支持 React、script-tag、CLI、拓扑 DAG 求解器和正式 JSON schema。

测试发现：All 343 tests pass, all packages build (core ESM+CJS+types, React wrapper, standalone IIFE+ESM bundles, and the playground static site), the CLI validates JSON and serves an interactive Gantt viewer on port 3456 returning HTTP 200. 14 分钟。

全能协议分析工具：内置浏览器抓包、MITM 代理、JS hooks、指纹伪装、AI 分析及 MCP server，用于 agent 集成。

测试发现：Installation, build, full test suite (185/189 passing), Electron binary, and built app all work in the container. Only expected dbus/GPU errors appear in container. 9 分钟。

用一条命令部署你的前端。支持 Claude Code Skills。

测试发现：PinMe CLI v2.0.11 builds and runs. 161 tests pass: 72 unit, 34 integration, 22 MJS regression, 31 CLI black-box, 2 npm pack. 16 分钟。

适用于 VS Code 的开源 AI 编码助手。Coding agent、代码补全、聊天、编辑和审查，支持本地或托管模型。你的模型，你的基础设施。

测试发现：npm install completed, TypeScript compilation and esbuild bundling succeed, the full VS Code extension test suite passes (642 passed, 0 failed), and both twinny-node and twinny-server CLI binaries respond correctly. 8 分钟。

支持 vim 键位绑定的代码审查 TUI

测试发现：tuicr builds, runs, and 1826 of 1832 tests pass. The binary responds to --version, --help, and review list subcommands correctly. 5 分钟。

一台你可以 curl 的计算机

测试发现：Open Terminal 0.14.0 is installed from source and running as a FastAPI/Uvicorn server on localhost:8768, with all tested API endpoints (/health, /execute, /files/list, /files/read, /files/write, /files/grep, /files/search, /files/delete, /system, /api/config, /skills) responding correctly. 8 分钟。

面向 AI agents 的隐身浏览器自动化，基于 MCP 和 Chrome DevTools Protocol：导航、网络钩子、DOM 提取和像素级准确的 UI 克隆。

测试发现：Stealth Browser MCP server runs on stdio transport, spawns a headless Chromium 157.0.8086.0, navigates to data URLs, extracts page HTML content, and closes cleanly -- all via MCP protocol over stdio. 15 分钟。

在本地跟踪查看器中拦截并检查来自 Claude Code、Codex CLI、Gemini CLI、Cursor CLI、OpenCode、Kimi/Kimi Code、Pi 和 Hermes 的 Coding Agent API 流量。

测试发现：Editable pip install succeeded, all 1136 tests pass, ruff lint and format are clean. 6 分钟。

PanWatch：覆盖 A股、港股和美股的 AI 股票监控，由 TradingAgents 提供支持。提供持仓分析、实时提醒和自动报告。

测试发现：Python 3.12 venv with all deps installed, 782/784 tests pass (3 skipped are Windows-only), server starts and responds 200 on /api/health, frontend builds cleanly via pnpm, WeasyPrint PDF renders Chinese characters as extractable text. 8 分钟。

适用于 Windows 和 macOS 的开源 AI agent 桌面应用，支持一键安装 Claude Code、MCP 工具和 Skills，并具备沙箱隔离、多模型支持以及 Feishu/Slack 集成。

测试发现：npm install completed successfully with Node.js v22.22.0. 155 of 158 test files pass (1141 of 1144 tests), MCP servers are built, better-sqlite3 native module rebuilt, and Electron v41.7.1 is ready. 63 分钟。

本地优先的 AI token 用量与成本追踪器，支持 31 款编码工具，包括 Claude Code、Codex、Cursor、Gemini 和 DeepSeek Harness，并提供原生应用。绝不读取提示词。

测试发现：All 3052 CLI tests pass with 0 failures; the tokentracker serve command starts and returns HTTP 200 on port 7680. 26 分钟。

面向 AI agent 和 IDE 的高性能代码智能引擎，支持 257 种语言、多仓库、基于图，可通过 CLI、MCP Server 和 API 访问。作为 AI coding agent 的队友，只暴露必要信息，将 token 使用量最多减少 50 倍。100% 本地。

测试发现：Go 1.27.0 installed locally. Gortex builds from source (v0.64.5) and the daemon indexes this repo into a knowledge graph of 182,637 nodes and 822,458 edges. Core graph, search, parser, resolver, query, analysis, semantic, serverstack, and agent test suites pass. The pre-built release binary v0.64.7 was also verified. 69 分钟。

UNIGIT 公共品牌与生态中心，AI 应该为每个人服务。

测试发现：npm install succeeds, npm run verify passes (22 files checked), all 4 media assets decode correctly via ffmpeg, and all tracked files are present on disk. 3 分钟。

28 个基于评估的心智模型和批判性思维技能，适用于 Claude Code、GitHub Copilot、Codex、Cursor 以及其他兼容 Agent Skills 的工具。

测试发现：All 314 tests pass, all 28 skills pass structural validation (100%), all 23 datasets pass split validation, and the judge reliability validator reports 100% agreement across 3 models. 6 分钟。

一个面向模型路由、多模态聊天、文件、工具、计费、身份和运营的企业级 AI 工作空间。

测试发现：The DEEIX Chat monorepo builds completely: Go backend compiles, all Go unit tests pass (internal/application, shared, ports, transport/http, etc.), Next.js frontend builds to static output, and the single binary serves both API and frontend on port 8080 returning 200 on health, API, and static endpoints. 23 分钟。

开源微信 Bot 管理平台 + App 应用市场 | 自托管微信 Bot 平台，带 App 应用市场 | 支持 Lark · Slack · Discord · DingTalk · GitHub · Notion · 20+ 应用 | AI 工具 | 7 种语言 SDK

测试发现：OpeniLink Hub binary built at /tmp/oih, launches on port 9800 with SQLite, serves the web UI, and all 14 Go test suites pass. 7 分钟。

Ouroboros, self-creating AI agent. Born Feb 16, 2026.

测试发现：Ouroboros 7.4.4 installed and builds; CLI launches, web server serves HTML UI on port 7777 (HTTP 200), default-lane pytest suite passes. 54 分钟。

AI Dictation App：开源且本地优先，打字速度快 3 倍，无需键盘。由开源模型驱动，离线可用，快速准确。

测试发现：Installation completed: Node 24.10.0, pnpm 10.34.5, all 1566 packages installed, whisper native addon compiled (.node binary at packages/whisper-wrapper/native/linux-x64/whisper.node), electron 44.3.0 available, workspace builds pass, and the test suite runs (149/150 test files pass, 1505/1508 tests pass, 3... 23 分钟。

Herdr 移动客户端，适用于 iPhone 和 Android。一个自托管的 PWA，可从手机驱动 Herdr、tmux 或 zellij 中的 Claude Code、Pi、Codex 和 OpenCode。推送提醒，无需应用商店。

测试发现：Collie builds and tests pass fully: CLI binary compiled at bin/collie (81 MB), PWA built into web/dist with index.html present, 555 combined tests across bridge, CLI, scripts, and web pass with 0 failures; the binary reports version 1.14.1 and doctor confirms a working install (no Herdr socket in this container, which... 12 分钟。

开源、隐私优先的客户之声分析，是一个自托管的 Enterpret 和 Unwrap 替代方案，在任何 LLM 看到反馈之前剥离其中的 PII。

测试发现：Obsei library installs, all 45 tests pass (including real model inference with transformers), and example scripts run successfully producing analysis output. 12 分钟。

测试发现：With a locally installed Node 22.23.3, `npm install`, the web build, and the desktop build succeed; the Electron desktop app launches on the virtual display with IPC handlers initialized and its SQLite database migrated (7 tables, 5 seeded skills), and the web app serves HTTP 200 responses with real HTML in both dev... 12 分钟。

免费开源的 Postiz 替代方案：自托管的社交媒体排程工具，配备 AI 帖子撰写器、内容日历，并支持跨 8 个平台发布。采用 MIT 许可证；外部提供商可能收取费用。

测试发现：The development server serves all pages (/, /login, /integrations, /gallery, /agents, /pricing) with HTTP 200 and returns proper 401 on protected API routes without a database or third-party credentials. 8 分钟。

一个 pod 用于你的团队和你各自运行的 agents：你自己的 Claude Code，或一键启动的托管 agent。开源（Apache-2.0），可自托管。

测试发现：Installed deps and built both workspaces, fixed the two real test-bug failures (expired fixture date, missing WebCrypto global), and verified the backend server live on localhost:5000: /api/health 200 healthy with real mongod connected, register 201, login 200 with JWT, while frontend's 962 Jest tests pass and full... 79 分钟。

根据你在 Jellyfin、Plex 或 Emby 上最近观看的内容，轻松向 Jellyseer/Overseer 请求推荐的电影、电视剧和动漫，让 SuggestArr 全自动处理，让你的媒体库保持新鲜，带来新颖刺激的内容。

测试发现：SuggestArr backend runs on Python 3.12, serves API on port 5000, initializes all database tables on first boot, and all 1050 pytest tests pass without errors. 3 分钟。

通过 SearXNG 为 AI 助手提供隐私网页搜索，支持 Claude、Cursor 和任何 MCP 客户端。

测试发现：The mcp-searxng 2.4.0 MCP server builds, passes all 885 tests, and serves valid MCP responses via both STDIO and HTTP transports. 3 分钟。

一款专为小说创作打造的跨平台、用户友好、AI Native 的一站式 Vibe Writing 工具。

测试发现：OpenFic v0.12.0 installs from source in a venv, all 1817 automated tests pass, the FastAPI server starts with SQLite database auto-migration, and its health endpoint responds 200. 11 分钟。

一个通过 ollama API 与本地 LLM 聊天的 Web 界面

测试发现：Ollama GUI v2.0.0 installs, builds, and passes its full test suite (88 unit + 38 e2e) in a container with Node 26 and bun 1.4.2. 4 分钟。

不在身边的时候，TA 还在这里。上传真实的聊天记录，重建数字人格：开源AI伴侣，具有人格蒸馏、情感引擎、11个LLM提供商和毕业机制。

测试发现：Mirrai 1.3.0 installs with pnpm, compiles with Vite and esbuild, passes 915 of 915 TypeScript + vitest test cases, and serves the full-stack app (React SPA + Express/trpc API + PostgreSQL 16) on port 3000 with admin auto-bootstrap, JWT authentication, login/register endpoints, and database connectivity confirmed. 15 分钟。

经过实战检验的 Claude Code、OpenAI Codex、Cursor 配置、插件、钩子和 agent，支持 Kimi、MiniMax 和 GLM API。

测试发现：The plugin validation script passes for all 34 plugins, version sync confirms all manifests are aligned, and the Astro static site builds correctly serving 55 pages with the homepage, plugin catalog, and sitemap all returning HTTP 200. 3 分钟。

开发者的第一个书签

测试发现：The static site builds from the canonical README (388 resources, 57 categories), all 49 unit and Playwright browser tests pass, both ko and en pages serve HTTP 200 at their project prefix with working 404 handling, and generated SVG/PNG assets decode correctly. 6 分钟。

同时运行 Claude Code、Codex 和 Gemini，每个都在自己的 git worktree 中

测试发现：On this container, with user-local Node 22.23.3: npm install succeeds, 4818 tests pass, npm run check is clean, the full build produces an AppImage and a .deb, and the packaged desktop app launches on the Xvfb display and keeps running while its renderer processes IPC. 17 分钟。

面向所有人的自我改进 AI 员工

测试发现：OpenCompany v0.2.1 installed from source: server starts on port 5678 with FastAPI app, health endpoint returns HTTP 200, Temporal worker pool connected, 186 node plugins loaded, 3787/3788 Python backend tests pass, and the built React SPA is served at the app URL. 9 分钟。

160-tool MCP server，通过免费公共 API 获取实时全球情报：市场、流动性、预测市场、SEC 文件、冲突、军事、飞机群早期预警、移民与边境流动、网络、气候、新闻等。包含带日期的观测历史、29 个研究提示、引用简报、地理围栏、SSE 仪表板。MIT 许可证，无需付费 API 密钥。

测试发现：Package installed in venv, all 900 mocked tests pass, CLI responds with 48 commands, MCP server starts with 132 registered tools, cache and circuit breaker are clean. 3 分钟。

在 Claude Code 上使用 Obsidian + Graphify，每次会话的 token 数量减少多达 71.5 倍。包含持久化记忆、代码库知识图谱和聊天导入管道。包含 PT-BR。

测试发现：All 6 CI checks pass (py_compile, ruff, JSON block validation, README parity, shellcheck), both Python scripts process real data producing correct Obsidian markdown with YAML frontmatter and auto-tags, the shell sync wrapper orchestrates the full pipeline, and the optional claude-conversation-extractor tool is... 5 分钟。

用于编程 CLI 的自托管 AI 网关：一个兼容 OpenAI/Claude/Gemini/Codex 的端点，带有多租户 Web 控制台、请求 logs 和消费配额。

测试发现：With user-compiled PostgreSQL 16.8 and Redis 7.2.5 and remote-management.disable-control-panel set, the CliRelay server (built from this repo) runs on 127.0.0.1:8317 answering /healthz 204, /v1/models 200 and /v1beta/models 200, and the full test suite passes when run sequentially (go test -p 1 -count=1 ./... exit 0... 37 分钟。

小说创作与角色扮演的一站式 AI 工作台。

测试发现：Denova v0.5.1 backend compiles and runs (version: dev), frontend builds and all 786 unit tests pass across 138 test files, all 7 Go root-module packages and the agent module pass both test and vet with no errors, and dependency integrity is verified via go.sum. 10 分钟。

开源、自托管的 AI 应用构建器：agent 在隔离沙箱中构建真实应用，运行于你自己的服务器上，每个应用都有预览 URL。一条命令即可自托管。MIT。

测试发现：Go 1.22.7 core test suite (23 packages) and console React/Vite unit test suite (33 tests) both pass; make build compiles sandboxd/runtimed Go binaries and bundles the console SPA; gofmt and go vet report no issues. 12 分钟。

自托管暗网 OSINT 平台。从查询到图表的自动化威胁情报，共 13 步。Recorded Future、DarkOwl 和 Flare 的免费替代方案。

测试发现：All 1294 pytest tests pass. VoidAccess 2.0.3 CLI installs and runs: version reports 2.0.3, status shows all subsystems operational, and investigate executes its pipeline (skipping LLM/Tor which are unavailable in this environment). 24 分钟。

jevos：Jev 和 Microsoft-Decision-1 的开源替代方案。支持是和否决策、选项和带概率的评分，可在笔记本电脑的 CPU 上运行。

测试发现：The jev project installs with uv, the llama.cpp runtime and a pinned GGUF model download with sha256 verification, and the HTTP server (port 8017) serves real noul decisions over /v1/systemone plus /health, /v1/models and /dino, all confirmed with curl status codes and JSON responses in this container. 7 分钟。

自托管的 AI SRE agent。你不必独自监控你的系统。Versus Incident 将跟踪你的服务、应用和系统，检测新的或意外的问题，并为你的团队提供调查所需的证据。

测试发现：The checkout now builds and runs: the live server answered 200 on /healthz and /, stored a 201-created webhook incident and returned it from /api/admin/incidents, and the full Go (40 ok) and UI (946 passed) test suites pass. 10 分钟。

TickDB：AI 原生实时股票 API 与金融行情数据 API，覆盖美股、港股、A 股、外汇、期货、指数、大宗商品、加密货币。支持 Skill、CLI、MCP、REST API 和 WebSocket。

测试发现：TickDB MCP server installs from source, all 49 unit tests pass, server starts on port 8000 with 13 MCP tools registered and responds correctly to MCP Streamable HTTP requests. 2 分钟。

一个用于递归 agent 的复合 agent 操作系统。同时也是 Grok Bot 和 Meta 的 Muse 的开源替代品。

测试发现：One non-root install with vendored Node v24.21.0 and pnpm 11.6.0: install, typecheck, build, and the full test:unit suite passed (two consecutive exit-0 runs), and the core server boots and serves real /api health/apps/agents/routines/action-executions responses on 127.0.0.1:4141. 54 分钟。

开源 coding agent 记忆。记录 issue、尝试、修复和决策，然后在 agent 重复已失败的方法之前发出警告。为 Claude Code、Cursor、Antigravity 和 Codex 提供原生 MCP server。100% 本地运行，无云端，无遥测。MIT 许可证。

测试发现：projectmem v0.3.3 installed in a virtual environment; the CLI tool, MCP server, test suite, and dashboard all run successfully with no errors. 2 分钟。

MCP-NixOS - 用于 NixOS 资源的模型上下文协议服务器

测试发现：All 3 NVF integration tests that were failing due to JS-chunked docs now pass, and the full test suite (495 tests) runs green; the server launches on STDIO. 8 分钟。

跨平台桌面应用，用于清理开发工件、管理本地开发工具以及检查常见环境问题。

测试发现：Frontend builds and serves (HTTP 200), TypeScript lint, release-version and AI-catalog validations pass, and the Rust core test suite passes 39/39; only the GTK/WebKit desktop shell cannot be compiled because the system dev libraries are missing and cannot be installed without root. 3 分钟。

从 Obsidian 中浏览、创建和编辑 AI agent 文件，支持 Claude Code、Cursor、Codex 及 12 种编码工具。

测试发现：In this container, the repo's full check gate passes: dependencies install cleanly with the frozen lockfile, the audit is clean, lint has 0 errors, all 24 tests pass, and the production esbuild build completes with no errors. 5 分钟。

自托管开源 AI 网关。一个端点支持 40 多个提供商，具备虚拟密钥、预算和使用量跟踪功能。

测试发现：The Otari LLM gateway installed from source with uv, boots as a standalone server on port 8000 serving the management API, bootstrap, health, docs, OpenAPI spec, and tutorial root page, and the OSS edition smoke test exercises all key lifecycle stages (health, bootstrap, user/keys, providers, routing, completions... 20 分钟。

拦截 LLM API 流量并在实时终端仪表盘中可视化 token 使用情况。跟踪成本、调试提示并监控 AI 开发会话中的上下文窗口使用情况。

测试发现：Tokentap is installed in a fresh venv, all 38 tests pass, and its proxy starts and correctly routes Anthropic and MiniMax requests to the real upstream APIs (verified by their 401 auth responses) while rejecting unknown paths with 400. 2 分钟。

TypeScript 样式指南和 Agent Skill。一套简洁的约定和最佳实践，用于编写一致、可维护的代码。

测试发现：Website builds, TypeScript compiles without errors, ESLint passes with zero warnings, development server launches and serves HTTP 200 on localhost, and the built site contains the TypeScript Style Guide introduction and table of contents with all 12 sections. 4 分钟。

Beads 的本地 UI，与你的 coding agent 协作处理 issues。

测试发现：Full dev environment working: Node 22 installed, all deps resolved, type checker and linter pass, 384 tests pass, server starts and serves the Beads UI over HTTP. 3 分钟。

基于 Rust 的代码智能 CLI，面向 AI coding agent。构建调用图和混合语义搜索索引（Dense + Sparse + RRF + Reranker），支持 7 种语言。作为原生 MCP 工具提供给 Claude Code 和 Codex CLI。

测试发现：codeseek 0.1.31 builds from source, all 104 tests pass, CLI commands (init, status, callers, callees, callgraph, snippet, skeleton, serve --mcp) work on Python/JS/Rust repos, and the npm wrapper auto-downloads the release binary for fresh users. 15 分钟。

开源自托管的 AI agent 运行时和多 agent 框架，用于自主 agent 群，提供 Agent 记忆、MCP 工具、调度、委派以及 23+ 个 LLM 提供商（Claude、GPT、Gemini、OpenRouter、Ollama），是 Claude Code 和 LangChain 的实用替代方案。

测试发现：SwarmClaw installs, builds, passes all 938 tests, and runs: the built and dev servers both answer 200 on /api/healthz, the web app serves HTML on /, the CLI answers --help and version 1.9.40, and the Electron desktop app launches on the virtual display and loads its UI. 9 分钟。

面向 AI agents 的 LLM 监督持久记忆：基于图的回忆、跨会话知识、单一二进制。兼容 DeepSeek Harness、Claude Code、OpenClaw 及任何 agent 运行时。

测试发现：All 19 deterministic Go test packages pass, the mnemon binary builds and runs, and the installed binary at $GOBIN/mnemon successfully executes memory operations (remember, recall, status) on a real SQLite-backed store. 5 分钟。

一个全面的 AI 开发模式集合，用于在 AI 辅助下构建软件，按实现成熟度和开发生命周期阶段组织。包含 Foundation、Development 和 Operations 模式，以及实际示例和反模式。

测试发现：All 556 non-slow tests pass on Python 3.12.3 with 22 pinned dependencies. 3 pre-existing slow tests fail due to external content drift (not fixable here). verification/STATUS.md regenerated. Cannot push to remote (no auth). 7 分钟。

最快的开源 AI Gateway 和 LLM Gateway，一个二进制文件内置控制台，用于路由、安全、预算和监控所有 LLM 调用，支持 OpenAI、Anthropic、Gemini、Bedrock 等。

测试发现：The Antirion AI gateway builds from source, its test suite passes 1044/1044 tests, the binary starts in standalone mode with the embedded store, and successfully routes a chat completion request through a provider, deployment, model, and key to a mock upstream and back with HTTP 200. 16 分钟。

一个由 AI 驱动的 README 和交互式 Wiki 生成器，适用于任何项目。AI 驱动的 README 及交互式 Wiki 生成工具，面向中文的开源 DeepWiki。

测试发现：The readmex skill is installed in the user's Codex skills directory and its scanner runs; a live codex run on a mock repo generated README.md, images/logo.svg, and 9 MkDocs pages, and the emitted site builds successfully with mkdocs 1.6.1 + material 9.7.7 from a local venv. 8 分钟。

将 AI 对话复制为干净的 Markdown 上下文包，一键从 ChatGPT、Claude、Gemini、DeepSeek、Grok 和 GitHub 复制。从侧边栏复制，无需打开。零上传，100% 本地，开源。

测试发现：CtxPort monorepo installs, builds (packages, Next.js web site with 25 static pages, Chrome MV3 extension bundle), passes all 30 unit tests, and its context-bundle to structured-markdown pipeline produces correct output verified end to end. 8 分钟。

<10 MB，自托管 AI 小说写作工具，单个 Go 二进制文件，带 Web UI。将创意转化为大纲和章节，支持审阅、伏笔、事实核查和全书润色。兼容 OpenAI API；支持中文和英文。

测试发现：The show-me-the-story Go binary builds (13 MB), serves the Svelte frontend on port 48090 returning HTTP 200, and the /api/status endpoint returns a valid JSON project status object. 5 分钟。

《小红书运营手册 · AI工作台》：用于 Xiaohongshu 运营的免费 Codex 技能

测试发现：All 7 skills pass validation, package correctly to .skill files, install into ~/.codex/skills/ and ~/.codex/skills/.system/, and are recognized by Codex CLI which loads the SKILL.md instructions and generates output matching the specified format. 2 分钟。

Claude Code 的本地分析仪表板。无需云端，无需遥测。

测试发现：cc-lens 0.4.1 installs, builds, and runs on Node 22 with all 54 tests passing and the demo dashboard serving HTTP 200 on all routes. 7 分钟。

CharacterArc（弧光）AI 小说创作应用，集成项目设定、角色关系、剧情大纲、章节写作与多模型 AI 协作功能。

测试发现：CharacterArc installs, builds with pnpm run build, passes all 187 tests via the Electron test runner with the added loader, and its desktop app launches on the virtual display, initializing a 37-table SQLite workspace database. 10 分钟。

开源的运行时 AI agent 安全工具：监控并控制 AI agent，在 agent 行动前实时捕获恶意工具使用、提示注入和策略漂移。

测试发现：Python SDK v1.2.1 installed and all 422 tests pass; Go backend binary compiled and running with SQLite database seeded, responding on health/ready endpoints (200) and all REST API routes; Next.js frontend dashboard built as standalone and starts successfully. 18 分钟。

为 Claude Code 赋予照片级记忆，只需一个可移植文件。无需数据库、SQLite 或 ChromaDB，仅需一个可提交到 git、通过 scp 传输或共享的.mv2 文件。原生 Rust 核心，操作耗时低于毫秒级。

测试发现：claude-brain builds with tsup, passes all 3 tests, hooks accept stdin and return valid Claude Code protocol, scripts create/search/query a portable .mv2 memory file, and included video assets decode without errors. 8 分钟。

开源、自托管的 AI 平台（PHP/Symfony）：可以与任何模型聊天，对你的文档进行 RAG，可嵌入的 widget，MCP：查看每个回答的成本。Apache-2.0。

测试发现：Frontend vitest unit test suite passes 3093/3095 tests (99.93%), 435/436 test files pass, with the remaining 2 failed tests being pre-existing ImpersonationBanner router navigation timing issues. 42 分钟。

面向 Genspark Claw、OpenClaw 与 Hermes agents 的社区技能包与 CLI：5 个生产就绪的 SKILL.md 技能（深度研究、报告、幻灯片、浏览器自动化、AI 通话准备）、零依赖的 Node.js 安装程序、MCP 服务器、一行安装

测试发现：Package installs with zero dependencies. All 12 tests pass. All CLI commands (list, doctor, validate, install, info, version, help) work correctly. MCP server responds to JSON-RPC tools/list and tools/call. Programmatic API via require('genspark-claw') exports listBundledSkills, findSkill, installSkills... 1 分钟。

在自己的电脑上构建 AI 电话 agent，并使用 AI 呼叫者测试每项更改。开源，无需 API 密钥即可免费运行，使用自己的密钥时每次通话只需几分钱。支持导入 Dialogflow CX。

测试发现：Both packages (vpd-vk 0.3.0, vpengine 0.3.0) installed in .venv-engine; 453 tests pass; engine starts and answers health and chat session endpoints. 23 分钟。

自主 AI 开发团队机器人

测试发现：Go 1.26.6 was installed manually, the module dependencies resolved, all 3 binaries build cleanly, all 34 test packages pass with the race detector enabled, and the go vet analysis reports no issues. 7 分钟。

人生K线：基于 AI 大模型 + 传统八字命理的人生运势可视化工具。将命运绘制成 K 线图，像看股票一样看人生。支持自托管部署。

测试发现：Server starts on port 3000, health endpoint returns 200, user registration and login work, profile CRUD works, local fortune calculation (daily/monthly/yearly) returns computed results, knowledge articles and cases are served, frontend builds cleanly, and the bazi calculator produces realistic 100-year K-line data. 7 分钟。

面向 Claude Code、Codex 和 Gemini 的自托管 AI API 网关：在一个端点后面聚合 API 密钥和订阅账户，支持故障转移、协议转换和成本跟踪。

测试发现：Binary builds, server starts and listens on :8080, health endpoint returns 200, admin login/auth flow works, all 49 Go test packages pass, all 284 JS frontend tests pass, model catalog syncs 244 models from models.dev. 10 分钟。

本地优先的 3 合 1 AI 记忆层和 MCP 服务器，适用于 Claude Code、Codex、Grok、Gemini、VS Code 和 Cursor。将会话历史、代码库索引和概念图融合在 SQLite 中。实现零云端、隐私优先的上下文和即时回忆，支持多 agent 集群。

测试发现：marm-mcp-server 2.48.0 installs in a venv, all 1370 non-skipped tests pass, the CLI prints version and help, and all 14 MCP tools are documented consistently across 9 surfaces. 57 分钟。

一个功能丰富的门户，通过轻量级 Node.js Web 应用与 GPT-4、Claude、Gemini、Mistral 和 OpenAI Assistant API 进行聊天；支持可定制的多模态，包括语音、图像和文件。

测试发现：Smoke test passes all 8 checks, server starts on port 3000 and responds 200 on /health and /api/system/status endpoints, all 9 provider handlers and 13 route groups initialize without errors. 7 分钟。

简历通 ResumeForge：本地优先的 AI 简历工作台，包含岗位管理、事实台账防虚构、AI 定制/手写简历、多格式导出、模拟 interview 与面试深挖、自动投递台、求职进度与数据看板。基于 FastAPI + React + SQLite，单机运行，数据不出本机。 | AI 简历构建器、求职跟踪器与...

测试发现：ResumeForge backend and frontend both install, build, and pass all 3602 tests (2801 backend + 801 frontend). The backend server starts, runs Alembic migrations on first launch, and serves the health endpoint at /api/health with status ok and version 0.15.0. 26 分钟。

一个开源、自托管的 AI 模型中心，兼容 Hugging Face，加速 vLLM/SGLang 性能。

测试发现：MatrixHub Go backend compiles and runs with SQLite on port 3001, health/endpoint returns 200, login (admin/changeme) authenticates and returns a session cookie, authenticated user CRUD works, built React UI assets are served correctly, and make test.unit passes all 17 test packages (excluding the known-broken hf... 11 分钟。

本地优先的二次元陪伴学习应用

测试发现：The backend API starts, all 549 Python tests pass, the Next.js frontend typechecks, lints, and builds, and the mock end-to-end flow (vault unlock, space creation, auto-assigned mock connections, text and streaming turn with valid structured responses, session end) works correctly without any real API keys. 7 分钟。

OpenAI Agents API，开源且可自托管。通过官方 OpenAI SDK，在您自己的机器上使用您自己的模型运行 Codex、Claude Code 或 MiniMax Code。

测试发现：OpenAgentCore development environment is fully operational inside this container: Go 1.26.10 compiler, PostgreSQL 16 (localhost:15432), Node.js 24.22.1, and pnpm 10.30.3 are all installed and working. All 5 Core binaries compile and produce executables. All 61 Go service packages pass their tests against a real... 49 分钟。

面向 CLIProxyAPI (CPA) 的自托管控制平面与用量可观测性控制台：AI 提供商、客户端密钥、流式遥测和模型成本快照。

测试发现：The binary builds, all Go packages test pass, the full frontend builds and embeds, the app serves the SPA and health endpoint with status ok in both demo mode (in-process fixture) and non-demo mode (with a mock CPA answering the v8 API probe). 16 分钟。

gpt-image-2 画图工作台 / 自托管的工作台，用于 AI 文案、海报、图像会话和视觉产品工作流

测试发现：Backend test suite 293/293 passing, web frontend builds to dist/ with vite, and the FastAPI server starts and returns HTTP 200 on /healthz. 8 分钟。

已归档；不再维护。

测试发现：All 35 backend test targets pass, the binary runs and serves /healthz → 200, /v1/models returns model list, /admin redirects to login, and the committed dist/ serves the dashboard; the frontend cannot be rebuilt from source due to Node 18 vs 20 requirement for Tailwind v4 but this does not affect the bundled binary. 6 分钟。

可自托管的 LLM 服务列表

测试发现：README.md regenerated from Harbor metadata with 161 LLM service entries, GitHub stats cached for 142 repos, Deno 2.9.8 installed. 8 分钟。

由 Docling 驱动的可视化文档分析工作室，配置提取管道，在浏览器中检查文本、表格和边界框，然后分块、嵌入并索引到 OpenSearch 和 Neo4j。

测试发现：Docling Studio backend serves /api/health (200), accepts file uploads, runs async Docling analysis pipeline to completion (COMPLETED status with rendered HTML and page metadata); pytest suite 854/869 pass; frontend vitest suite 432/432 pass; frontend builds to dist/. 11 分钟。

本地优先的个人操作系统：情绪、睡眠、习惯、任务、日计划以及将一切串联的记忆。兼容 Ollama、OpenAI、Anthropic 和 OpenRouter。

测试发现：Installed and built in a venv; full test suite passes 2327/2327; all CLI entry points respond to --help; package imports as hermes_life_os version 1.40.0. 2 分钟。

OpenScribe 是一个开源的 AI 记录员，能够自动记录患者就诊并生成结构化临床笔记。你可以完全掌控数据、工作流程和患者隐私，无供应商锁定。

测试发现：The monorepo installs, TypeScript compiles, and 260 unit tests pass across all packages. The Next.js web app builds and serves HTTP 200. Three source files were patched to fix crypto availability in Node --test workers. 16 分钟。

免费开源的离线写作助手，完全在本地协助语法、重写与语气。

测试发现：Backend API responds on /health, all 11 backend tests pass, all 213 frontend tests pass, frontend builds and dev server starts. 3 分钟。

开源、自托管的 AI agent 平台，用于你公司的 AI 队友。在自有服务器和模型上构建、运行和治理 agent：共享聊天、RAG、技能、真实浏览器、4,700+ MCP 服务器、预算、审批和审计追踪。是 ChatGPT Enterprise 和 Grok Bot 的替代方案。Apache-2.0，基于...

测试发现：Backend Python environment (Python 3.15.2 via uv) installed, 254 packages resolved; frontend Node dependencies (771 packages) installed via bun; backend lint passes (ruff, formatting); 609 backend unit tests pass covering coverage gate, agent registry, labels, A2A client/server, mutation locks, observability, and... 15 分钟。

驱动 Flowbaker 的执行引擎，一个用于构建、连接和自动化工作流的无代码平台。

测试发现：Flowbaker Executor builds from source with Go 1.27.3, its test suite passes entirely (3 packages, 142 assertions), and the binary responds to all CLI commands (help, status, start). 4 分钟。

这并非又一个 ChatGPT 包装器，而是一个本地优先的工具，可将你的笔记、网页、PDF 和音视频转化为个人知识图谱，完全在你的机器上运行。

测试发现：Backend API serves 30 endpoints with seed graph data, all 131 automated tests pass, and the React frontend builds and starts. 4 分钟。

一个自托管的 AI 助手，用于学术研究：管理你的论文、搜索文献、跟踪截止日期，并在你已使用的渠道上回答你。

测试发现：Package installs cleanly via pip, all 52 existing tests pass, the CLI shows all commands (agent, gateway, doctor, onboard, stop, login, reset), the gateway server starts and serves the Web UI on port 18790 responding HTTP 200. 7 分钟。

人们付费的 SaaS 工具的开源替代品

测试发现：The catalog validator runs on the bundled Python 3.12.3 and reports 'Validated 13 catalog entries.' with exit 0, all catalog JSON parses, and the working tree stays unchanged. 1 分钟。

免费、开源替代 ChatGPT Spaces：一个可自托管的 AI 工作空间，用于共享页面和对话。

测试发现：API server on port 8000 and Next.js client on port 3000 both serve HTTP 200 responses; the API returns/creates spaces, pages, messages, and auth records correctly over a local SQLite database. 10 分钟。

开源 AI coding agent，专为大型项目和真实世界任务而设计。

测试发现：Plandex server starts on port 8099 with PostgreSQL 16.4 and LiteLLM proxy on port 4000; server tests pass; CLI builds and authenticates against local server; full CRUD API for projects, plans, branches, settings is operational. 18 分钟。

可私有部署的多租户知识智能体平台：统一 RAG、知识图谱、多智能体、MCP/Skills、沙盒与权限管理。Yuxi = Cloud Agents + Knowledge RAG，一个自托管的知识 agent 平台，用于 RAG、知识图谱和多 agent 工作流。

测试发现：Backend Python dependencies installed and locked; all 1640 backend unit tests pass; all 164 web unit tests pass; Vite production build completes successfully; engineering trust contracts and lint checks pass. Two minor test permission-assertion bugs fixed (umask compatibility) and one Node.js version gap resolved... 25 分钟。

自托管 AI 会计应用。用于收据、发票、交易的 LLM 分析器，支持自定义提示词和分类。

测试发现：TaxHacker Next.js app builds, all 33 tests pass, dev server runs on port 7331 and serves the TaxHacker UI with PostgreSQL 16 connected and all 10 Prisma migrations applied, creating 14 database tables. 12 分钟。

面向生产环境 LLM 和 agent 系统的 AI 可观测性平台。

测试发现：The logfire SDK version 5.0.0 is installed, its CLI reports version info, and 47 of 48 test modules pass with ~1392 tests passing, 2 skipped, and 0 failures. 45 分钟。

在浏览器中审查 AI coding agents 的计划、差异和运行中的应用。本地优先，可与任何 agent 配合使用。

测试发现：Built crit binary compiles, launches an HTTP server on 127.0.0.1, serves /api/health returning 200, auto-detects git changes in feature branches, serves embedded frontend, and the full test suite passes all unit and JS tests except one pre-existing intermittent flake in session lazy-threshold ordering. 42 分钟。

HarnessRouter Community Edition：自托管、Apache-2.0 版的 agent 工具统一接口。通过一个 API 运行 Codex、Claude Code、Hermes、PI、DSH 等，支持会话、流式传输、文件、取消和失败处理。实现了开放标准 Unified Harness...

测试发现：Gateway and runner servers start and respond to health checks; all three test suites (gateway: 622/17, runner: 490/2, protocol conformance: 85/2) pass. 6 分钟。

freehire：开源求职搜索引擎

测试发现：Postgres 17 (pgvector), Redis 7.4, Meilisearch 1.55, the Go API server on :8080, and the built SvelteKit frontend on :4173 are running; the API answered 200 on /health and /api/v1/jobs/search with 717 live Greenhouse jobs, and the web app served /jobs with real job links. 46 分钟。

一个为 Hermes Agent 设计的 7 层内存操作系统：持久化内存（使用 Qdrant）、结构化事实、结构召回、自动整理的知识库以及精准上下文注入。可在本地运行，支持任何 LLM 提供商。

测试发现：Python dependencies install cleanly in a venv, all 7 icarus module imports succeed, all 3 standalone test suites pass with 0 failures, SQLite databases are initialized with correct FTS5 schema, Icarus plugin tools produce valid output, and the Docker-compose file, worker services, and cron scripts are syntactically... 8 分钟。

Vibe AStock：A股短线复盘与跟踪工作台 · 盯盘异动、情绪复盘（连板天梯、题材轮动、明日验证）、AI 复盘与周报、交易日志、盘前提醒，182 个数据端点。支持 Claude Code、Codex、WorkBuddy 订阅或任意模型 API。 | 一个用于中国 A 股的短线交易复盘与跟踪工作台。

测试发现：Python venv installed, Python and Node dependencies satisfied, frontend built, server starts and responds on both health endpoints, all 1282 backend tests pass, all 36 frontend tests pass. 22 分钟。

本地优先的 AI 驱动的文档智能平台，用于调查新闻

测试发现：The SHATTERED API server runs on http://127.0.0.1:8100 with 26 shards, 559 API endpoints, PostgreSQL+pgvector backend, serving a healthy health-check and Swagger UI. All core services (database, config, storage, chunks, vectors, events, workers, models) report active. 23 分钟。

低代码工具，用于快速构建和协调多 agent 团队

测试发现：Frontend installs and builds (vite dev server serves 200), and the FastAPI backend installs with Poetry, connects to a real local PostgreSQL 18.4 and Redis 8.10.2, applies Alembic migrations, and serves authenticated API requests (login returns a JWT, GET /users/me returns the admin user) while its pytest suite... 21 分钟。

开源 DeepWiki：面向 GitHub/Gitlab/Bitbucket 仓库的 AI 驱动 Wiki 生成器。

测试发现：Backend FastAPI server runs on port 8001 with health, models, auth, repo, wiki, chat, codemap, and system endpoints all responding; frontend Next.js dev server compiles and serves on port 3000; 100 out of 105 tests pass (2 deselected for real API keys, 3 stale/invalid tests). 21 分钟。

React.js 的拖放式邮件编辑器组件

测试发现：The react-email-editor package installs, builds (tsup), and all 12 unit tests pass under Node 22 with --experimental-require-module flag. 4 分钟。

Telegram 中的 AI 助手，能记住一切并帮助你管理生活。一条命令即可自托管。

测试发现：Iva v0.4.8 dependencies installed and CLI launches (bin/iva.mjs --help shows all commands); test suite runs at 3646/3649 passing, with three pre-existing test-environment failures: OverlayFS container metadata edge case, EPIPE on Claude subprocess mock pipe, and cross-test module cache pollution in card-store resolver. 64 分钟。

免费和低成本 AI 工具、LLM APIs、IDEs、agent 和基础设施的精选列表，用于构建真实 AI 应用

测试发现：The Next.js website serves every route (/, /news, /search, /categories, /stacks, /privacy, /terms, /submit, /api, and dynamic /tools/[id], /stacks/[id], /categories/[slug]) with HTTP 200 and the /api/news endpoint returns live scraped news items from news.smol.ai, with the production server staying up under repeated... 3 分钟。

本地优先的开源 AI 助手，用于管理你的数据。统一管理任务、笔记、文档、照片和书签。私密、自托管，并通过 API 可扩展。

测试发现：On this container, Eclaire installs and builds (pnpm install, app:upgrade to 0.7.0 on SQLite, pnpm build all packages, typecheck 16/16), 2185 unit tests pass across all six packages, the backend serves HTTP 200 on /health and the SPA, the CLI status/settings/user commands work against the live backend, and the... 58 分钟。

用于 Web 自动化的开源 AI agent。使用你真实且已登录的浏览器来自动化日常任务。

测试发现：Dependencies installed, project builds to dist/ (background.iife.js 1,848.81 kB, manifest.json), TypeScript type-check passes across all 15 packages, unit tests pass (14/14). 15 分钟。

AutoClip｜一个链接，一键出片。开源 AI 视频剪辑桌面工具，将播客、访谈、课程等长视频自动剪成短视频，生成字幕、封面和发布文案，适配抖音、小红书、TikTok、Reels 与 YouTube Shorts。开源 AI 视频剪辑与内容再利用。

测试发现：The AutoClip FastAPI backend serves all API health endpoints and the test suite passes with 103/103 tests passing. The frontend builds to production with Vite. 4 分钟。

用于自托管 LLM 工具调用和多步骤 agent 工作流的 Python 框架

测试发现：forge-guardrails 0.9.5 installed with dev dependencies; 1571 unit tests pass; core, proxy, guardrails, and client modules all import cleanly. 6 分钟。

面向个体创业者的 Agent 技能

测试发现：The repo installs via npx skills add with Node 22 and every reddit skill script now fetches live Reddit data (posts, search, comments, users, subreddits), while the website worker serves 200 HTML, skills.json (10 skills), and robots.txt; logo and banner crop scripts produce images that ffmpeg decodes without errors. 24 分钟。

由 AI 驱动的 StartUp Accelerator Engine，使用 Next.js、LangChain、PostgreSQL 和 pgvector 构建。上传、整理并与文档聊天。包括预测性缺失文档检测、基于角色的工作流和页面级洞察提取。

测试发现：LaunchStack monorepo installs cleanly after Node upgrade to v20.11.0; package-level vitest (327/327) and web jest (hundreds of unit+integration tests) all pass against a standalone PostgreSQL 16 with pgvector; only missing real external credentials for chat, S3, and provider APIs. 33 分钟。

用于 agent 协调的开放标准

测试发现：Cotal 0.100.2 installed, pnpm build succeeded across 29 packages, and the local mesh runs with broker (nats-server v2.14.0), delivery daemon, and manager all in 'running' state on static auth, verified by 27 passing smoke suites covering multicast/unicast/anycast delivery, DM durability, presence, auth, the cotal-lang... 10 分钟。

Agent-ready RPA 套件，提供开箱即用的自动化工具。为个人和企业打造。

测试发现：Engine Python 3.13.15 dependencies installed, engine modules import correctly, 207 pytest tests passing across data processing and encryption components, 1 source bug fixed (string fill slice type). Frontend shared and CLI packages build. Can't run backend services or Electron desktop app without Windows/Docker. 22 分钟。

在你自己的 PC 上 24/7 工作的自主公司：Windows/Linux/macOS。

测试发现：The dashboard and product-center HTTP servers, the Python test suite (619 of 632 passing; 13 isolation-blocked failures), all fake-engine shell integration tests, frontend JS contract suites and example Node tests except scopefence run successfully in this container. 18 分钟。

面向 AI agent 的 Trace-native CI/CD：生产故障转化为阻止 PR 的回归测试。自动检测、聚类、冻结为封闭用例，并在 CI 中以 $0 成本重放。

测试发现：Python backend and SDK install, import, and all 1174 pytest tests pass on Python 3.12.3. Frontend installs via pnpm with no errors and next build compiles successfully but OOMs during static page generation (container memory limit). vitest tests require Node 22+ ([email protected] and [email protected]) while this... 43 分钟。

一个 Spring AI 库，为你的 AI 应用带来受 Claude Code 启发的工具和 agent 技能。

测试发现：The core library spring-ai-agent-utils and all submodules build and all 495 tests pass (0 failures, 0 errors, 3 platform skips) with JDK 17 and Maven 3.9.8. 10 分钟。

开源 Python 库，用于 AI 浏览器 agents，以及 Browser Use Cloud：隐身浏览器和托管 agent API。

测试发现：browser-use v0.13.10 installed in a Python 3.12 venv with uv, Chromium headless installed via playwright, and all 563 CI tests pass (529 passed, 34 skipped) after fixing 2 test identity comparison bugs in test_beta_agent.py that only manifest under xdist worker parallelism. 75 分钟。

将任何网站变成 CLI，并通过 AI agent 使用你已登录的浏览器。

测试发现：OpenCLI 1.8.8 builds and runs on Node 22.14.0; 632 test files pass (7325 tests, 3 skipped) across unit, extension, adapter, smoke projects; CLI binary answers version 1.8.8 and lists 1366 registered commands. 27 分钟。

离线隔离、注重隐私的开源 NotebookLM 替代方案。一个 AI agent，用于研究、转换和编辑你的文档。

测试发现：All Python trees (surfsense_backend, surfsense_mcp, surfsense_local/backend, plugins/remote/host core tests) install and pass tests; the web frontend fails to compile native deps on Node 18.x (requires 24.x); docs are clean. 20 分钟。

使用 pi 进行氛围编程，在一个轻量级、真正的 IDE 中，它会根据你的工作方式自我定制：你需要的氛围

测试发现：890 deps installed, 14 packages typecheck, 14/14 test suites pass (987 in server alone), web app builds, E2E 183/183 passes, server /health returns 200. 18 分钟。

一个个人知识库，能够自我构建和维护。放入来源，Claude（或 Codex/Gemini）读取它们，提取知识，并维护一个持久的相互链接的 wiki。适用于 Claude Code、Codex、OpenCode、Gemini CLI。无需 API 密钥。

测试发现：llm-wiki-agent repo has 9 wiki pages (1 source, 2 entities, 5 concepts, 1 overview), a working knowledge graph (9 nodes, 36 edges), and all 10 Python tools run without errors after fixing 3 bugs in tools/_utils.py and tools/ingest.py. 20 分钟。

生成干净的 2D 游戏精灵与动画图集。组件行流程：状态行、alpha 清理、帧提取、运行时图集。Codex/Claude 技能。

测试发现：sprite-gen 2.11.0 installed in Python 3.12 venv with dev deps via pip; Pillow 12.3, NumPy 2.5 resolve; CLI responds with full help text; img2webp 1.5.0 with -exact support on PATH; 2020 tests all pass within 300s timeout. 75 分钟。

一个在您自己的 PC（Windows/Linux/macOS）上 24/7 运行的自主公司。

测试发现：All 419 Python tests pass, all 4 shell test suites pass, all 19 frontend JS tests pass, dashboard server starts and serves HTTP 200 with structured JSON on / and /api/status. 9 分钟。

在 Android 上通过一条命令运行 OpenClaw，无需 proot，无需 Linux

测试发现：OpenClaw 2026.9.8 installed via npm; gateway starts cleanly and serves OpenClaw Control HTML dashboard on port 19000 responding HTTP 200; installation verification passes all 12 applicable checks. 7 分钟。

桌面 AI agent 会为每项任务雇用专家，能在您的真实应用中持续工作数周，并接受来自手机的指令。免费，Apache-2.0。

测试发现：Agentlas OS runtime v1.2.57 installed to ~/.agentlas/runtime/current with hephaestus doctor ok; ontology runtime operational with SQLite storage, Model2Vec int8 embeddings, FTS adapters, and auto-ingest working from ~/.agentlas/inbox directories. 6 分钟。

租用模型，拥有其他一切。免费、开源的 AI 应用，用于实际工作。

测试发现：The Cowork OS repo installs and builds on Node 24, the full vitest suite passes (855 files, 8918 tests), and the cowork CLI answers with version 0.5.54. 71 分钟。

开源个人理财应用，内置 AI 财务顾问：在您自己的电脑上追踪净资产、投资、ETF、现金和债务。

测试发现：The Aurelio backend starts, serves its API on port 8001, answers health checks, and returns demo data from its survey, institutions, and dashboard endpoints; all 977 backend tests pass. 5 分钟。

clodfarm（大声读出来）：一个 Claude Code agent 农场。种下一个任务，它们会将其拆分为 sub-agents，打开工作，并根据每个账户真实的 5 小时和每周使用量来调节节奏。从 Claude app 中操控它。

测试发现：clodfarm 1.13.0 installed and its full test suite passes (411/522 tests, 111 skipped due to missing Claude Code binary); CLI responds with version and help output; SQLite and DynamoDB store backends are functional. 30 分钟。

功能完整的 AI Agent 桌面客户端，支持 Webui 访问，并可创造性地定制和扩展。

测试发现：The Go gateway binary (agent-gateway / cmd/gateway) builds, starts on port 50052, and returns 200 on /api/status with proper Bearer auth; all Go tests pass; Node script tests (30/30), virtual-core tests (34/34), and GUI release backend tests pass; the desktop GUI (Tauri) cannot build due to missing system libraries... 32 分钟。

本地模型上的私有 AI 应用的完整 API 层：RAG、技能、工具、MCP、text-to-sql 等。可与任何 OpenAI 兼容的推理服务器配合使用。

测试发现：PrivateGPT server launches with mock LLM/embedding models on port 18080, responds to /v1/models, /health, and /v1/messages API endpoints. Test suite: 2079 passed, 1 failed (pre-existing), 4 errors (transient DB state). 51 分钟。

基于 Playwright 和AI实现的闲鱼多任务实时/定时监控与智能分析系统，配备了功能完善的后台管理UI。帮助用户从闲鱼海量商品中，找到心仪产品。

测试发现：Python backend with 117 passing tests and API server running on port 8000; frontend Vue build blocked by Node 18.19.1 < 20.19 requirement. 7 分钟。

ODS V3 预发布版：在正式 V3 发布前进行公开测试和优化。将你的 PC、Mac 或 Linux 设备变成私有 AI 服务器。

测试发现：Project builds correctly, all 418 bats tests, 14/14 smoke tests (after ai_err fix), ~300 shell/Python contract tests, and the vite frontend build pass. The Docker runtime install cannot complete in this environment, but the dry-run installer, CLI (-v2.6.0), all standalone test suites, and the frontend are functional. 15 分钟。

自托管的 TypeScript agent 运行时，具备持久审批和可验证的运行记录。拥有它，批准它，审计它。

测试发现：The monorepo installs, builds, passes 2350 unit tests across all 4 workspaces plus the benchmark harness, and the CLI prints help, all with Node.js 22, no API keys or network required. 4 分钟。

别再切换多个 AI 账户了。Quotio 是一款精美的原生 macOS 菜单栏应用，可统一管理你的 Claude、Gemini、OpenAI、Qwen 和 Antigravity 订阅，并提供实时配额跟踪和智能自动故障转移，适用于 Claude Code、OpenCode 和 Droid 等 AI 编码工具。

测试发现：quotio CLI 1.0.0-beta.4 installs and runs on Linux, its full Rust test suite passes 634/634, and `serve --no-saved-accounts` returns HTTP 200 schema-2 JSON on /v2/snapshot; macOS/iOS apps were not runnable in this Linux container. 14 分钟。

开源 AI 人性化工具：将 AI 生成的文本重写为自然、像人写的文字。LLM 重写 + 多引擎翻译，带有逐步追踪。Python CLI、n8n 和 Docker。

测试发现：Project installs in a venv, all 40 unit tests pass, CLI entry point works, and real OpenRouter LLM rewriting (Steps 1-2 of the pipeline) is verified end-to-end. 10 分钟。

一个用 Rust 编写的安全持久化个人 agent 服务器。单个二进制文件，沙箱执行，multi-provider LLMs，语音，记忆，Telegram，WhatsApp，Discord，Teams 和 MCP 工具。安全设计，运行在你的硬件上。

测试发现：Moltis gateway binary compiles from source in a fresh Ubuntu 24.04 container, launches on HTTP, prints the setup code banner, and 2810 unit tests pass across 15 of 59 workspace crates, with 5 pre-existing infrastructure-missing test failures (ssh-keygen, WASM, Docker). 48 分钟。

用于下一代语音 agent 的基础设施，旨在提供通用记忆。它分为左脑和右脑，分别存储信息和情感，而完全流式架构从根本上消除了延迟。

测试发现：VoiceMem 0.2.3 installed in Python 3.12 virtual environment at /tmp/venv; left-brain memory ingest/search pipeline functional with mock OpenAI API; 5/5 unit tests pass; audio perception models not downloaded (ASR/VAD warmup skipped gracefully); PortAudio-dependent capture unavailable (no root). 15 分钟。

需求分析效率提升 200%，全球首个为 AI 编程时代设计的团队协作 MCP 服务器，自动分析需求并自动编写前后端代码，下载切图。

测试发现：Lanhu MCP Server 1.8.6 installed, all 216 tests pass, the server launches on HTTP port and responds with HTTP 200 to proper MCP JSON-RPC initialize requests. 7 分钟。

使用一条命令在 Android 上运行 OpenClaw，无需 proot，无需 Linux

测试发现：openclaw 2026.9.8 npm package installed, gateway running on ws://127.0.0.1:19099 responds HTTP 200 with live status, CLI version/status/doctor commands all functional. 5 分钟。

自然语言驱动的无限画布工作流 Agent，让 AI video 创作第一次真正变成可编辑的工作流。 AICON 面向创作者，提供从剧本拆解、分镜生成、素材生成、video 合成到内容分发的一整套能力。 不是只给你一个输入框，而是让你用自然语言和无限画布一起驱动创作，把文本、图片、video...

测试发现：Backend FastAPI server starts and serves root, health, and API v1 endpoints on port 8000. Frontend builds to dist/. 57 backend unit tests and 15 frontend utility tests pass. 22 分钟。

用于 Metal、CUDA 和 Vulkan 的 LLM 推理引擎。

测试发现：tensorfold 0.6.0 installed and its pure-Python test suite passes on x86_64 Linux; the CLI shows version 0.6.0 and lists all supported model families; the server module imports cleanly. 29 分钟。

自托管网络搜索，适用于AI agents和SEO。Google, Bing, Yandex, Baidu, DuckDuckGo & Ecosia，在真实浏览器中渲染：排序结果和干净的页面Markdown。API, CLI, MCP

测试发现：OpenSERP v0.8.13 builds, all 12 unit test suites pass, and the HTTP server starts, serving /health and /ready endpoints on 127.0.0.1:7000. 8 分钟。

开源的可观测性工具，使用 AI agents 自动修复你的软件

测试发现：pnpm install succeeded (51s). Node v22.23.3 with pnpm 9.15.9. 143 proxy tests pass, 78 worker tests pass, ~380+ total tests pass across all packages with 0 failures. Full stack requires Docker for postgres/clickhouse/collector which is unavailable without root. 46 分钟。

WebAI2API: 基于 Camoufox 的网页 AI 转 OpenAI API 工具，支持 LMArena/Gemini 等，多窗口并发与账号隔离。

测试发现：The WebAI2API server starts on port 3000 with a Camoufox browser instance, serving OpenAI-compatible /v1/models, WebUI, and /v1/cookies endpoints, all returning HTTP 200, though chat completions require real AI website credentials via the browser adapter and cannot produce output without logging in. 7 分钟。

一个基于源的 Meta Muse 集成和社区连接器技能目录，包含能力、认证和权限说明。

测试发现：This is a community-maintained catalog of 150 connector skill folders (SKILL.md + Python CLI) for Meta Muse and 12 workflow templates. All Python scripts have valid syntax, accept --help, handle invalid args gracefully, and their credential plumbing (bearer header, query param, path segment, inline surrogate) works... 7 分钟。

Prismer 云

测试发现：TypeScript SDK (built, 569 offline tests pass), Python SDK (installed, 149 offline tests pass), MCP Server (built, 12 tests pass), OpenCode Plugin (built), AIP SDK TypeScript (built, 17/17 compliance tests pass), Runtime (built) -- all 6 TS+Python packages build and pass their offline unit tests. 9 integration tests... 20 分钟。

不在身边的时候，TA 还在这里。上传真实聊天记录，重建数字人格，开源 AI 伴侣，具备人格蒸馏、情感引擎、11 个 LLM 提供商以及毕业机制。

测试发现：Installed Node 20 + pnpm, embedded PostgreSQL 16, .env with JWT secret; pnpm test (6/6), build, and tsc check pass; production server answers 200 and a full register-login-persona-upload-analyze-chat flow completes against a local OpenAI-compatible mock. 8 分钟。

全网最全的、持续更新的、最火爆的 100+ 人格蒸馏技能合集｜多 agent 系统｜名人/导演/天涯大神/古籍/二次元/职场/情感全品类

测试发现：34 SKILL.md files with valid YAML frontmatter across 5 sections: DirectorAgents-1, EQ, Novelists, tianya-gods-framework, zimeiti. Python orchestrator produces correct multi-agent decision reports. All shell scripts pass syntax checks and execute correctly. 1 分钟。

为你的 AI agent 提供一个真正的浏览器，并让人参与其中。开源 MCP 原生浏览器 agent。

测试发现：Controller and client packages install in a venv; 1189 test suite tests pass via FastAPI TestClient with mocked browser node; Docker-based compose stack not available so full end-to-end with Playwright cannot run. 6 分钟。

一个自托管的 OpenCode 命令中心，将会话、git、终端和日程集成在一个 Web 应用中，可通过手机或桌面访问。

测试发现：Dependencies installed, backend builds and serves HTTP on :5003 returning 200, test suite passes 3081 of 3148 tests (145 of 161 files) with 13 failures caused by missing external tools (unzip, opencode binary, msb) and API-key-dependent services. 33 分钟。

Claude Code Skill Factory 是一个强大的开源工具包，用于大规模构建和部署生产就绪的 Claude Skills、Code Agents、自定义 Slash Commands 和 LLM Prompts。轻松生成结构化的技能模板，自动化工作流集成，并通过简洁、对开发者友好的设置加速 AI...

测试发现：All 61 Python modules compile, 14 JSON files validate, and all 6 skill generator tools execute end-to-end with bundled sample data: hook-factory, prompt-factory (7/7 validation), scrum-master-agent (5 stories parsed, sprint metrics computed), slash-command-factory (17 presets), agent-factory (valid YAML)... 23 分钟。

利用 LLM 进行代码补全（以及更多？）的 LSP 服务器

测试发现：Workspace builds with rustup-managed Rust 1.99.0, all 13 unit tests pass, and llm-ls 0.5.4 serves initialize/getCompletions/acceptCompletion/rejectCompletion successfully against the schema-correct local mock server. 22 分钟。

开源技能，可学习任何品牌并将其转化为完整的设计系统。适用于 claude code 和 codex。一次安装，你的助手构建的每个 UI 都符合你的品牌。

测试发现：hue skill installed under ~/.codex/skills/hue and its validation gate passes on all 17 bundled examples plus a freshly generated contract-correct skill folder (No findings, exit 0). 3 分钟。

不是问答机器人，而是能真正干活的数字分身。共享智能体，共享项目上下文，让项目所有成员一起协同工作。同时智能体会像人一样通过资料或者工作抽象和总结经验到项目上下文中

测试发现：Backend API runs on port 8002 (SQLite, auth-bypass) with /health 200, project+knowledge APIs working, and a real agent turn streamed an SSE-mocked response end-to-end; 7 agent-core sandbox/shell tests remain red in this container due to missing root for python symlink/user-site and venv path layout. 37 分钟。

PiG (Pi in Go) 是对上游 Pi（即 Pi coding agent 背后的 TypeScript 代码库）的忠实 Go 移植。它是一个保持行为一致的翻译，而非重写：上游行为是约定，Go 是实现语言。

测试发现：pig binary builds and runs (0.4.1+1.0.3), all importable Go packages compile and vet cleanly, 70+ test packages pass, Node subprocess tests work after upgrading Node to v22. Full make test-blocked by tmux/cargo/rustc (no root). 34 分钟。

开源端到端 GEO 实现：状态分析、诊断、策略、工单、执行、验证

测试发现：Dependencies installed in a Python venv; all 131 unit tests pass; the GeoLook dashboard starts on port 8765 and serves HTTP 200 with the complete single-page UI; CLI subcommands (init, list, ui, --help) all work. 3 分钟。

使组织能够创建和部署可定制的 AI agent，支持多个 LLM 提供商、实时聊天、语音、文档训练、分析、用户管理、SSO 认证和可嵌入的小部件。基于 React 19、Redux Toolkit、Radix UI 和 Tailwind CSS 构建。

测试发现：pnpm install completed 1905 packages without error, pnpm test passes all 12480 unit tests, and next dev serves the root page at localhost:3000 with HTTP 200. 14 分钟。

给 agent 一台机器，但不是你的。每个 AI coding agent 都有自己的隔离机器，具备 root、Docker 和 systemd 权限，主动防御会自动检测并阻止威胁。

测试发现：The coi binary builds from source and runs all available Go unit tests successfully (23 packages); the binary produces correct help output and version string; 115 of 140 CLI Python integration tests pass (the 4 failures and 4 teardown errors all require a real Incus daemon). 13 分钟。

给 agent 一台机器。但不是你的。每个 AI coding agent 都有自己的隔离机器，带有 root、Docker 和 systemd，主动防御会自动检测并阻止威胁。

测试发现：The coi binary builds and runs (version prints), 23 test packages pass in short mode, go vet is clean for project packages, and all build-time dependencies (Go 1.26, libsystemd headers, nft) are available via user-local paths. 8 分钟。

AI 驱动的轻量级 RSS 阅读器。支持：GitHub Pages | Vercel | Alibaba Cloud ESA Pages | Docker

测试发现：FeedMe RSS aggregator installs, builds, and serves via Vite (HTTP 200 on both dev and production preview), and the data pipeline fetches 30 real RSS sources and generates summaries via a mock LLM API. 6 分钟。

50+ 个开源生成式 AI 应用，可克隆、部署并变现，包括图像生成器、视频工具、虚拟试穿、AI SaaS 模板和平台集成。每个模板均支持一键 Vercel 部署。

测试发现：social-post builds and serves 5/5 routes with HTTP 200 (static pages) or HTTP 401 (auth-required API) using SQLite database and mocked third-party credentials. 15 分钟。

通过 MCP 读取、写入、搜索并精准编辑 Obsidian 仓库中的笔记、标签和 frontmatter。支持 STDIO 或 Streamable HTTP。

测试发现：Dependencies installed, project builds, all 2608 tests pass, and the server starts via stdio with 14 tools and 3 resources registered. 2 分钟。

使用 Lynkr 简化您的工作流程，这是一款 CLI 工具，可作为 HTTP 代理，通过 Claude Code CLI 实现高效的代码交互。

测试发现：Server starts, responds to health checks with 200, and 1555 unit/integration tests pass across routing, compression, memory, caching, rate-limiting, and telemetry modules with mock credentials. 31 分钟。

用于 Coolify 的 MCP 服务器，提供 42 个优化工具，通过 AI 助手管理自托管 PaaS

测试发现：npm install and tsc build succeed; 1050 unit tests pass; the MCP server initializes over stdio with correct capabilities; doctor CLI verifies all checks against a mock Coolify instance. 11 分钟。

结合 chrome-devtools-mcp 的能力与 Skill 规范，实现基于 JSRPC、Flask 和 autoDecoder 方案的前端 JS 逆向自动化分析，提升 JS 逆向效率。

测试发现：The project installs cleanly in a venv, all 22 Python/JS scripts run successfully, the Flask proxy generates and starts, and validate_artifacts passes all schema/syntax/cross-file checks, only browser-dependent features (Chrome, JSRPC live server, real target website) remain untested in this container. 4 分钟。

一项面向 AI agent 的技能：使用 SearXNG 搜索网络，使用 Camofox 浏览网页，使用 CloakBrowser 绕过保护。设计上防幻觉。自托管、免费、无限制。

测试发现：npm install succeeds, CloakBrowser (with Node 20, not the system Node 18) fetches live pages, the SearXNG and Camofox deterministic scripts pass their full command flows against schema-correct mocks, unit suites pass 28/28 and 12/12, and setup.sh/audit.sh complete without failed checks; real SearXNG/Camofox Docker... 33 分钟。

基于 Rust 的开源 LLM 网关。一个端点即可接入 OpenAI、Anthropic、Gemini 及其他所有供应商。任何 SDK、任何模型、任何工具。

测试发现：respan-gateway 0.3.0 Python package installed and verified working against a mock provider; Rust binary v0.2.0 built from source and all 692 unit tests pass. 10 分钟。

注重细节的 AI 聊天应用。所有顶级模型、真正的记忆、活文档，以及你自己的工具。

测试发现：All 3 test suites pass (239 tests total): 106 app unit tests, 67 backend unit tests, and 66 backend integration tests. The Next.js production build compiles successfully with all 26 static pages pre-rendered. The dev server starts and serves a 200 response on / with 91KB of app HTML. Pages through Clerk's middleware... 23 分钟。

镜子中的自托管个性化 AI

测试发现：MirrorMate installs and builds, its 27-test suite passes, and the built server serves all major pages plus config, memory, and mock-Ollama-backed chat endpoints with verified 200 responses and SQLite persistence on this Node 22 + rebuilt better-sqlite3 setup. 10 分钟。

AI 每日新闻简报 · GitHub 热门 + X 热门文章 + 行情技术分析 · 23 个数据源聚合 + LLM 中文摘要 · 本地或 GitHub Actions 部署

测试发现：Full pipeline runs end-to-end: 549 articles fetched from 23 sources, LLM enrichment mocked with deterministic JSON for all 5 enrichment categories + trading commentary + daily digest, report written to daily_reports/2026-10-23 as JSON+HTML+articles cache. 34 分钟。

在 Claude Code CLI 中无缝轮换 Claude 和 GPT 的订阅及 API 密钥，在达到使用限制时切换而不中断会话，全部通过本地仪表盘控制。100% 本地。

测试发现：The 1.3.1 daemon installs, starts on 127.0.0.1, serves health/profiles APIs, gates the proxy path from unauthenticated callers, and passes its entire 1696-test suite with zero failures on Linux/Python 3.12. 11 分钟。

用于 LLM 基础设施的高性能轻量级代理和负载均衡器。支持智能路由、自动故障转移以及跨本地和远程推理后端的统一模型发现。

测试发现：Olla v0.0.x builds from source with Go 1.24.13, starts on localhost:40114, serves health endpoint returning healthy, passes all 42 test packages, validates configuration, accepts and responds to HTTP requests, and passes the make ready pre-commit gate (test + fmt + vet). 6 分钟。

自托管的音乐发现工具，适用于你的音乐库。发现艺术家和专辑，查看推荐，并将其发送到 Lidarr、slskd 或播放列表。自带 AI 提供商或本地模型。无需 Lidarr 也可使用。

测试发现：digarr 1.18.0 is installed with Bun 1.4.4, dependencies are resolved, TypeScript type-checking passes, lint passes with no errors, the Vite frontend and Bun backend both build, embedded PGlite migrations apply on startup, the server boots and responds to HTTP requests, and 3550 of 3567 tests pass (15... 8 分钟。

将你的 ChatGPT 或 Claude 账户变成一个完全可用的 OpenAI 兼容 API。无需 API 密钥。支持工具调用、视觉、文件附件和图像生成。

测试发现：CatGPT Gateway starts with PROVIDER=minimax, passes all 7 MiniMax provider tests, serves health/endpoints/models/chat completions on port 8000, and responds correctly to OpenAI SDK and curl requests. 6 分钟。

为任何内容聘请一位私人 AI 家教：它了解你实际的学习方式，并专门为你教授下一课。Bloom's 2-Sigma 研究作为 Claude Code 技能 + 可自托管的 Web 应用 · 中文优先 苏格拉底式 AI 家教

测试发现：Backend 49/49 tests pass, frontend 46/46 tests pass, backend launches and responds on port 8000, frontend builds to static dist, all with mocked LLM calls (no real API key needed). 3 分钟。

轻量级自托管 AI agent。开源 OpenClaw 替代方案，使用 Go 编写。

测试发现：GoGogot binary builds, all 4 Go module tests pass, and the agent initializes and polls for updates against a mock Telegram server with a mock LLM provider configured. 9 分钟。

Pharos：本地优先的 agentic RAG，用于团队文档库：多格式导入、混合检索、企业级 ACL、双 HTTP + MCP 出口。

测试发现：Pharos installs, builds, and runs its test suite (240/240 pass in the product+engine CPU tests); the HTTP server starts and answers health checks; the MCP stdio adapter initializes with full tool contract; the RAG pipeline (embedder → Qdrant in-memory → hybrid dense+BM25 retrieval → small-to-big context assembly)... 22 分钟。

面向语音、聊天和视频的开源 AI 面试平台

测试发现：npm install completes, project builds with next build, dev server starts on port 3000 serving login/register at HTTP 200, 350 web tests pass, and 13 Playwright functional tests pass. 11 分钟。

一个开源的、云原生的、高性能的网关，统一了多个 LLM 提供商，从本地解决方案（如 Ollama）到主要云提供商（如 OpenAI、Groq、Cohere、Anthropic、Cloudflare 和 DeepSeek）。

测试发现：Inference Gateway builds from source, installs via install.sh (writable dir), and serves /health 200, /v1/models 200, streaming and non-streaming chat 200 against a mock OpenAI-compatible upstream; the full go test -race suite passes (211 tests, 0 failures). 12 分钟。

开源桌面应用，用于本地 coding agent。查看每个工具调用、diff 和终端命令。自带模型；会话留在你的机器上。

测试发现：The openagentd 2.26.0 backend server and web UI build cleanly, the suite passes 4728/4728 (12 skipped) plus 3168 web tests, the server answers 200 on /api/health/ready and /api/agents and /api/agents/registry, and a real chat turn completed against a local OpenAI-compatible mock, saving the assistant replies into the... 20 分钟。

轻量级 AI 原生工作流构建器，面向个人和小型团队：用自然语言描述你的想法，即可在可视化画布上获得可运行的工作流，并将其发布为 API。一条命令即可通过 Docker 自托管。是 Dify / n8n 的开源替代品。

测试发现：Install completes, Next.js dev server starts on port 5000 and returns 200 on /api/health, and the full test suite of 765 tests across 68 files passes cleanly. 5 分钟。

一个开源的统一接口，用于运行和管理自托管的 LLM。

测试发现：After fixing the Alembic downgrade revisions, OpenGateLLM installs, its unit (580) and integration (747 pass, 1 skipped) suites pass against real local Postgres 16.2 and Redis 6.2.14, and uvicorn serves /health 200, bootstraps the admin, and completes a real chat request end to end through a local mock provider. 30 分钟。

Xiryn：开源 personal agent。能在你自己的计算机上完成 Muse、Grok Bot 和 Dots 所做的一切：每个 agent 在 Docker 中拥有自己的 Linux 桌面，任意模型配合你的密钥，一个加密保管库，以及通过 Telegram 进行控制。

测试发现：Xiryn v0.5.1 on Next.js 16.4.3 with Node.js 22.13.0: npm install, build, typecheck, lint, and format all pass; the web app starts and serves all routes (200) with a correctly initialized SQLite database (12 tables) and a mock API key in .env; Docker is unavailable in this container so agent workstation features could... 6 分钟。

一款强大的浏览器扩展，为 DeepSeek Chat 提供自定义工具、持久化内存和无缝项目脚手架功能。支持生成和下载文件、在浏览器中运行 Python、创建 PowerPoint 和 Excel 文档、构建多文件项目为 ZIP 压缩包，以及管理自定义技能和角色。

测试发现：npm install and npm run build succeed; five verified test lanes pass here (npm test exit 0: 1608 unit, 36 Chrome E2E, 4 Firefox E2E, plus 21 Android-simulator E2E), and MCP tool access works against a locally stood-up schema-correct mock server. 28 分钟。

AI 获取上下文，而非你的隐私数据。面向浏览器聊天、AI API 和编码 agent 的本地优先隐私代理。

测试发现：PasteGuard 0.9.3 runs from source in this container: /health returned 200 (detector up, local_llm up), /info and /dashboard returned 200, /api/mask masked PII and secrets into placeholders, and /openai/v1/chat/completions returned 200 for both upstream-forwarded plain text and local-routed PII/secret requests against... 38 分钟。

开源 AI agent 上下文层。PipesHub 将公司的知识（Slack、Drive、Jira、GitHub、Microsoft 365 及 50 多个连接器）转化为支持权限感知的工作区，agent 可以搜索、grep、导航和引用。支持 MCP、SDK 和内置 agent。可自托管。

测试发现：286 Python unit tests pass, 9633 JavaScript unit tests pass (4 pending). Both suites execute against mocked external services without real infrastructure dependencies. 62 分钟。

停止配置你的 AI 技术栈，开始使用它。一条命令即可带来一个完整的预配置 LLM 技术栈，包含数百个可供探索的服务。

测试发现：Harbor CLI v0.5.6 installed from source, frontend web app builds, all Deno-based unit tests (41 total) pass, lint self-test green (12 bash rules, 9 compose rules, 1 boost rule, 3 orchestrator checks). Docker-dependent container test matrix blocked. 16 分钟。

一个无代码 RAG 服务器，让答案与你的文档保持同步。支持 8 种向量数据库。基于 Pathway 构建。

测试发现：All 394 non-integration tests pass under the mock license patch, including DuckDB-backed end-to-end indexer tests with persistence, the combined serviette up server+indexer, and the FastAPI server. 86 分钟。

将 Windsurf / Devin Desktop 的 100+ 个 AI 模型（Claude、GPT、Gemini、DeepSeek、Kimi、GLM、SWE）转换为兼容 OpenAI、Anthropic 和 Gemini 的 API。面向 Claude Code、Cline 和 Cursor...

测试发现：The Node.js HTTP server starts on port 3003 and all 4317 tests pass; the Language Server binary is unavailable in this container (requires Windsurf IDE binary), so the proxy functions as a Devin Connect gateway rather than a full Cascade/LanguageServer bridge, which is the expected state when DEVIN_CONNECT is not... 36 分钟。

隐私优先的 AI 会议纪要工具，适用于 Windows 和 macOS。免费、开源（MIT）。设备端 Whisper/Parakeet 转录和说话人分离，无会议机器人。开源替代品，替代 Granola、Otter 和 Fireflies。

测试发现：Meetily Tauri app builds from source, all Rust (188) and frontend (8) tests pass, and the desktop app launches successfully on Xvfb, initializing all subsystems including Tauri GUI, notification system, database, tray icon, and model managers. 61 分钟。

开源的 Terraform Enterprise 替代方案

测试发现：1318 Python unit/service/runner/auth/crypto/storage tests pass (out of ~6200 total); lint and format clean; python-hcl2 pinned to 7.3.1; two source files fixed for TYPE_CHECKING annotation guards; 12 test files updated for starlette 1.3 _IncludedRouter compatibility. 42 分钟。

本地私有、开源的自进化跨平台 AI 内容发现 agent：先理解你，再主动从 B站、小红书、抖音、YouTube、X、知乎、Reddit、微博等平台与开放 Web 寻找内容。（支持 deepseek harness 插件）

测试发现：OpenBiliClaw v0.3.218 is fully installed, builds, passes lint, and 1007+ tests pass. CLI commands work. The API server starts and responds to health/init-status endpoints. 76 分钟。

一个使用 Next.js、Tailwind CSS 和 TypeScript 构建的现代化 AI agents 聊天界面。

测试发现：Build and typecheck pass cleanly, production server answers HTTP 200 on all routes. 11 分钟。

Plan-Apply-Unify 循环：为 Claude Code 设计的结构化 AI 辅助开发方法。质量优先，而非为速度而速度。

测试发现：The PAUL Framework installer deploys 97 markdown files (28 commands, 27 templates, 23 workflows, 14 references, 5 rules) to a .claude/ directory with correct path replacements, across local, global, and custom-config install modes; the --help flag prints usage; all error cases exit with proper error messages. 3 分钟。

基于 X-Bit 量化的设备端 LLM 推理

测试发现：C demos build and run from command-line, Python binding installs and loads the native library, dialog formatting works for all 11 model types, but full model inference requires a Picovoice AccessKey and proprietary .pllm model files not present in this container. 14 分钟。

开源、自托管的企业级和站点搜索服务器，基于 OpenSearch 构建。可爬取 Web/文件/数据库/云来源，支持 20+ 种语言，提供 REST API，并具备 AI/RAG 及语义搜索功能。Apache-2.0。

测试发现：Fess 15.9.0-SNAPSHOT builds and packages successfully from source. All 7,634 unit tests pass. The embedded Tomcat binds port 8080 but the application fails at runtime because no OpenSearch backend is reachable on localhost:9200. 36 分钟。

面向 AI 原生公司的自托管 AI Agents 平台。Claude Code、Codex、Gemini、OpenCode agents。Apache 2.0。

测试发现：无法验证；log 显示了它停在哪里。 29 分钟。

一个命令安装器，用于在自有服务器上自托管 AI 技术栈：n8n、Ollama、Open WebUI、OpenClaw、Dify、Flowise、Supabase、ComfyUI、Qdrant 及 30 多种工具。基于 Docker Compose，通过 Caddy 自动配置 HTTPS，内置监控。免费的...

测试发现：The repository is structurally valid (Docker Compose YAML, shell scripts, JSON configs, Python files) but cannot install or launch because Docker is absent and cannot be installed without root privileges in this container. 13 分钟。

Krawl 是一个可定制的轻量级云原生 Web 诱骗服务器与反爬虫工具，通过逼真的随机生成诱饵数据和 AI 生成的 HTML 模板，创建带有明显漏洞的虚假 Web 应用。

测试发现：Krawl standalone mode builds and runs: after pip install -r requirements.txt, uvicorn app:app --app-dir src serves HTTP 200 on /dashf/healthz, answers the honeypot with 204 on /, records POST credentials from 198.51.100.77 into SQLite, and the dashboard API shows that IP categorized as attacker. 92 分钟。

自托管 AI 网关，用于私有 RAG、自然语言数据访问和工具调用 agent。

测试发现：无法验证；log 显示了它停在哪里。 28 分钟。