created 2026-10-13 · seen 3h ago
Local error/fix memory for Claude Code — records repeated errors + their fixes in SQLite, injects the fix before you re-debug. Offline, deduped, secret-redacted.
New open-source developer tools, the day they appear.
Most tools worth adopting surface in the noise before anyone writes about them. GH Radar sweeps newly created repositories every day and keeps the ones that look like real developer tooling, so you can filter the flood down to the handful actually worth a look.
2,814 matching
created 2026-10-13 · seen 3h ago
Local error/fix memory for Claude Code — records repeated errors + their fixes in SQLite, injects the fix before you re-debug. Offline, deduped, secret-redacted.
created 2026-10-13 · seen 3h ago
面向中文生态的 AI Agent 应用底座:Mastra 内核 + 六家国产大模型适配 + 微信/企微渠道接入 + 合规护栏,全部为可插拔 adapter。
created 2026-10-13 · seen 3h ago
AI-native data warehouse workbench: chat-to-SQL, auto-modeling coding agent, built-in scheduler, knowledge graph & governance — with enterprise-grade security. 中英双语 | 仓颉 AI 数仓工作台
created 2026-10-13 · seen 3h ago
Local decision engine with typed, calibrated outputs (Noul/Choice/Score): cheap→strong cascade, ECE/Brier calibration, SystemOne protocol interop. Apache-2.0.
created 2026-10-13 · seen 3h ago
Keep your AI coding agents on track. Railspike leads to clear requirements, plans you approve, and atomic implementation stages with verification and a commit trail. Plain markdown, easy to install, tested extensively with Cursor, Windsurf, Copilot, Codex, Claude Code, and others.
created 2026-10-13 · seen 3h ago
Serverless OpenAI-compatible LLM on vast.ai spot GPUs — auto-launch on call, auto-destroy on idle, fixed URL
created 2026-10-13 · seen 3h ago
AI video creation workbench (open source)
created 2026-10-13 · seen 3h ago
并行隔离多智能体执行引擎:零互知、三道容错门、可注入时钟的确定性测试(零运行时依赖)
created 2026-10-13 · seen 3h ago
Remote control your local Windows Codex from ChatGPT via a secure Bridge/Tunnel/MCP connection.
created 2026-10-13 · seen 3h ago
AI code review for regulated codebases, built in n8n: redaction before the LLM call, human approval in Slack, and an insert-only audit log. The model drafts; a person decides.
created 2026-10-13 · seen 3h ago
Open-source browser video editor with a first-class MCP server: AI agents edit the timeline while you watch live.
created 2026-10-13 · seen 3h ago
A terminal client for the chat server, keeping server text separate from what you are typing.
created 2026-10-13 · seen 3h ago
A small, self-contained Slack app for weekly office attendance planning.
created 2026-10-13 · seen 3h ago
麦麦护栏:把麦当劳账号交给 AI 之前,先装个刹车。本地 MCP 安全层:下单等写操作需本机确认、消费上限、只读模式、审计日志,Token 只在本机。Human-in-the-loop guard proxy for McDonald's China MCP.
created 2026-10-13 · seen 3h ago
Your backlog as markdown files. Ranked, versioned, agent-ready
created 2026-10-13 · seen 3h ago
分流镜:看清 OpenClash / mihomo 每条连接来自哪台设备、属于哪个服务、走了哪条规则和节点,并用内核真实判定任意域名、IP、App 的分流
created 2026-10-13 · seen 3h ago
🚪 Many services, one front door
created 2026-10-13 · seen 3h ago
两款轻量级 Python 信息收集工具:Kiss Scan 子域名爆破(1581 条字典 + 标题抓取 + ICP 备案)+ 端口扫描(top1000 + HTTP 指纹 + 774 条服务字典)
created 2026-10-13 · seen 3h ago
react agent component for an instant local first agent in web
created 2026-10-13 · seen 3h ago
Parallel question branches for the Pi coding agent — fork your session, interrogate several questions at once in side-sessions, merge back only the distilled conclusions. Keeps your main context clean.
created 2026-10-13 · seen 3h ago
Full-stack invoice processing: Tesseract OCR, OpenCV preprocessing, field extraction and LLM-powered search (FastAPI + React)
created 2026-10-13 · seen 3h ago
Pack a repository into a single, LLM-friendly Markdown file — tree, files, and a real token budget.
created 2026-10-13 · seen 3h ago
Keeps Claude Code's browser automation and Figma MCP calls cheap, with a Figma call counter and status-bar count.
created 2026-10-13 · seen 3h ago
A CLI that opens a local GitHub pull request dashboard. Review incoming PRs, track your own, and submit reviews from your browser.
Every repo carries a potential score combining log-scaled stars and forks, recent commit activity, project health, and how deeply it matched the developer-tool topics the collector watches. Stars and forks are log-scaled so a single outlier cannot flatten everything below it, and forks carry their own weight, so a young project people are already building on can outrank a more starred one nobody has adopted.
Age is deliberately excluded. Every repo in the collection is currently the same age, so a growth rate would carry no information and would just rank by raw stars. The weighting is worth revisiting once the collector has covered enough days for velocity to mean something.
Muhammad Uzair Shahid Rao
Senior Software Engineer & Ecosystem Architect
Lahore, Punjab, Pakistan
I build tools that make other people's work faster, and I ship them in public while I do it.
Most of them started as friction of my own: a component library I needed twice, an automation I was tired of running by hand. They turned out to be useful to people I'll never meet. GH Radar is the same bet. I wanted a way to find what was worth adopting before the noise buried it, so I built one and left it here for anyone else doing the same hunt.
The goal isn't a list of repos. It's a body of open, dependable tools that other developers reach for without thinking twice.