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.
7,357 matching
created 2026-10-15 · seen 3d ago
Retrieval-augmented Q&A over your docs: hybrid pgvector + Postgres full-text search with reciprocal rank fusion, validated citations, retrieval evals gating CI, and an MCP server. Python, FastAPI, LangChain splitters, local embeddings.
created 2026-10-15 · seen 3d ago
Shrink coding-agent context: trim tool output + session notes. CLI & MCP server. Defaults to AtmoRouter, a cheap AI API gateway and OpenRouter alternative.
created 2026-10-15 · seen 3d ago
Token-saving OpenAI/Anthropic-compatible proxy with $ savings report. Defaults to AtmoRouter, a cheap AI API gateway and OpenRouter alternative.
created 2026-10-15 · seen 3d ago
Persistent cross-session memory for coding agents (Go, MCP, SQLite FTS5). Defaults to AtmoRouter, a cheap AI API gateway and OpenRouter alternative.
created 2026-10-15 · seen 3d ago
A.I. first project management in real time
created 2026-10-15 · seen 3d ago
WooCommerce crypto payment gateway plugin. Accept USDT, USDC, BTC, ETH, SOL, TRX and 200+ coins straight to your own wallets. No KYB, no custody.
created 2026-10-15 · seen 3d ago
Self-hosted publishing platform for manga and webcomics.
created 2026-10-15 · seen 3d ago
Push-based heartbeat + tokens/sec monitor for self-hosted LLMs (Ollama, llama.cpp, vLLM). Discord alerts, no open ports. Built and run by an AI agent.
created 2026-10-15 · seen 3d ago
Which Claude model should each agent use? A source-checked field guide to Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, with eight installable Claude Code subagents, a 24-task selector and copy-paste workflows. Independent guide, not an Anthropic publication.
created 2026-10-15 · seen 3d ago
Self-hosted dashboard showing real-time and historical Google Analytics 4 data for all your sites on one screen. Plain PHP, no dependencies.
created 2026-10-15 · seen 3d ago
Own your tokens: Qwen3.8-27B-Flash-MTP - a 27B ternary model on an 8 GB laptop GPU at 70 t/s. Launchers, features.ini and a chat client. Binaries on Releases, weights on Hugging Face.
created 2026-10-15 · seen 3d ago
Unofficial CLI for Tongji Yunpan (AnyShare): upload, download, and long-term share links
created 2026-10-15 · seen 3d ago
Let Codex or Claude Code talk to your signed-in ChatGPT: send project updates, get a second opinion, keep one ongoing conversation. Local MCP server + Chrome extension, PDF support, no API key. Unofficial; AI-generated with Codex and Claude Code.
created 2026-10-15 · seen 3d ago
Learn decision models by building with them. TypeSafe Jev and Microsoft Decision-1 route support tickets in under a second for about $0.02 per 1,000 calls; Claude handles the hard cases. Interactive site, benchmark, calibration tests and docs.
created 2026-10-15 · seen 3d ago
Official MCP server for Nifty: connect Claude, ChatGPT, Cursor and other AI agents to your Nifty workspace.
created 2026-10-15 · seen 3d ago
Verification gate for AI coding agents: blocks the agent from finishing while its changes introduce typecheck, lint or test failures. Repo map, context, impact and probes over MCP.
created 2026-10-15 · seen 3d ago
Run repeatable agent experiments in Docker: capture inputs, track tool calls, token usage and cost, grade outputs, and compare baseline vs candidate with evidence for every result.
created 2026-10-15 · seen 3d ago
mise for MCP servers: the right servers and accounts per directory, for every AI agent (Claude Code, Codex)
created 2026-10-15 · seen 3d ago
Local coding-agent experiments comparing continuation, fresh context, and branching.
created 2026-10-15 · seen 3d ago
Write your tasks in TypeScript
created 2026-10-15 · seen 3d ago
A byte-pair-encoding tokenizer in Lean 4 with machine-checked proofs: lossless round trip, never expands, no out-of-range token ids.
created 2026-10-15 · seen 3d ago
Run builds and renders through a shared slot and free-memory gate, so parallel AI agent sessions start heavy work a few at a time. Windows, Linux, macOS.
created 2026-10-15 · seen 3d ago
Ubuntu Linux lab covering systems administration, permissions, processes, services, networking, storage, Bash automation, and incident investigation.
created 2026-10-15 · seen 3d ago
Offline, deterministic grader for LLM chat answers: fixed questions, safety rules, reproducible score (Deno/TypeScript)
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.