Auto-curated daily. A scheduled job reads github.com/trending ↗, the best available Groq model picks the genuinely new + on-topic repos (LLM serving, agents, RAG, training/eval, dev tools, models, infra), and the change is committed — so every addition is dated on What’s new. The model only classifies; repo name, stars & description come straight from GitHub, and any pick it can’t match to a real trending repo is discarded. Capped at 8 new/day, deduped, never an empty commit.
▶ Run now triggers the same job on demand. It’s once-per-day: whichever run comes first — your click or the schedule — claims the day; the other is a no-op. So on days you don’t run it, the schedule does.
⚠ The daily job hasn’t added anything for 8 days (last run 2026-10-15). Check the Bitbucket schedule (Repository → Pipelines → Schedules) and its GROQ_API_KEY / BB_PUSH_TOKEN secured variables.
Trending now — 41 curated repos · last run 2026-10-15 · catch-up · Claude workflow, GitHub-API-verified
| Repo | Category | Lang | Stars | Gained | Why it’s here | Ingested |
|---|---|---|---|---|---|---|
MCP server + CLI that lets coding agents reverse engineer binaries, apps and sites | Dev tools | TypeScript | 266,053 | +32,282 | Agents can inspect native binaries (via Hopper/Ghidra/IDA), Electron apps, APKs and websites without the source code. ⚠ Star count unverified: under ~1k until early Oct, ~215k this week, 25% fork ratio; stargazers could not be sampled | catch-up · API-verified 2026-10-15 |
Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codeb | Agents | Python | 23,067 | +1,876 | Trending Python · 1,876 stars today | heuristic seed 2026-07-20 |
AI gateway | Serving | TypeScript | 25,494 | +1,651 | Unified API for 500+ models | Groq · llama-3.3-70b-versatile 2026-07-23 |
ADE for agents | Agents | TypeScript | 26,327 | +1,271 | Run coding agents with subscriptions | Groq · llama-3.3-70b-versatile 2026-07-23 |
AI agent toolkit | Agents | TypeScript | 75,646 | +919 | Unified LLM API and agent loop | Groq · llama-3.3-70b-versatile 2026-07-23 |
Rust headless browser kernel (V8 + Servo/Stylo) for agents; CLI, CDP and WebDriver | Dev tools | Rust | 18,651 | +6,415 this week | Fetches pages as Markdown or a semantic tree; vendor benchmark: 81.9% task success vs Chrome 99.8%, far less memory ⚠ Vendor benchmark: 81.9% task success vs Chrome 99.8%; repo is ~2 months old | catch-up · API-verified 2026-10-15 |
MCP server plus hooks that keep raw tool output out of coding agents' context windows | Dev tools | TypeScript | 29,714 | +902 | Sandboxed exec and SQLite/FTS5 session continuity across compaction for 17 coding agents; the 98% figure is one 315KB-to-5.4KB example. ⚠ Elastic License v2 (source-available, not OSI); the 98% figure is one README example | catch-up · API-verified 2026-10-15 |
Self-hostable agent memory server with retain/recall/reflect APIs, backed by PostgreSQL | RAG / memory | Python | 50,013 | +26,274 this month | Memory layer for stateful agents: 25+ LLM providers, Python/TS/Go clients, MCP; claims LongMemEval SOTA with cited outside reproduction. ⚠ Accuracy and "Fortune 500" claims are the vendor's own; upsells a hosted cloud | catch-up · API-verified 2026-10-15 |
The open-source AI voice studio. Clone, dictate, create. | Other | TypeScript | 44,132 | +839 | Trending TypeScript · 839 stars today | heuristic seed 2026-07-20 |
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas | Agents | Shell | 134,682 | +744 | Trending Shell · 744 stars today | heuristic seed 2026-07-20 |
The go-to web for your AI coding agent — local-first search, fetch, crawl & research over | Agents | TypeScript | 2,505 | +695 | Trending TypeScript · 695 stars today | heuristic seed 2026-07-20 |
The most intelligent agent harness for code | Agents | Rust | 9,597 | +612 | Trending Rust · 612 stars today | heuristic seed 2026-07-20 |
Text-space optimizer | Training & eval | Python | 14,512 | +599 | Trains reusable NL skills | Groq · llama-3.3-70b-versatile 2026-07-23 |
Self-hosted router pooling free tiers of 34 LLM providers behind one OpenAI-style /v1 API | Serving | TypeScript | 35,687 | +583 | Live example of rate-limit-aware routing with 429/5xx failover and key rotation; also serves Anthropic and Gemini APIs. ⚠ "Personal experimentation only"; pooling free tiers may breach provider terms | catch-up · API-verified 2026-10-15 |
Plugin that captures coding-agent sessions, summarizes them, and injects context later | RAG / memory | TypeScript | 100,983 | +3,371 this week | Adds cross-session memory to Claude Code, Codex, OpenCode and more via SQLite + Chroma search ⚠ Default installer opts you into a hosted, paid observer; pass --provider to stay local | catch-up · API-verified 2026-10-15 |
A Flexible Framework for Experiencing Heterogeneous LLM Inference/Fine-tune Optimizations | Training & eval | Python | 18,724 | +448 | Trending Python · 448 stars today | heuristic seed 2026-07-20 |
Kimi Code CLI is your next CLI agent. | Agents | Python | 10,220 | +405 | Trending Python · 405 stars today | heuristic seed 2026-07-20 |
ggml speech-to-text inference for 16+ model families | Models | C++ | 1,225 | +401 | Trending C++ · 401 stars today | heuristic seed 2026-07-20 |
AI Agent Assistant & development framework that integrates lots of IM platforms, LLMs, plu | Agents | Python | 37,066 | +330 | Trending Python · 330 stars today | heuristic seed 2026-07-20 |
CLI proxy for LLMs | Dev tools | Rust | 72,663 | +305 | Reduces token consumption | Groq · llama-3.3-70b-versatile 2026-07-23 |
Use claude code, codex or pi for free from the terminal, IDE, or you phone like OpenClaw ( | Dev tools | Python | 41,229 | +303 | Trending Python · 303 stars today | heuristic seed 2026-07-20 |
AirLLM 70B inference with single 4GB GPU | Infra | Jupyter Notebook | 23,823 | +283 | Trending Jupyter Notebook · 283 stars today | heuristic seed 2026-07-20 |
Very low latency speech to text, intent recognition, and text to speech, for building voic | Agents | C++ | 9,806 | +264 | Trending C++ · 264 stars today | heuristic seed 2026-07-20 |
Sandboxed runtime that enforces per-agent file, syscall and network policies | Infra | Rust | 16,328 | +7,698 this month | Kernel-level isolation, credentials injected only for approved endpoints, formal checks on policy changes; Docker/Podman/K8s ⚠ Telemetry on by default (OPENSHELL_TELEMETRY_ENABLED=false); pre-1.0 | catch-up · API-verified 2026-10-15 |
Use Garry Tan's exact Claude Code setup: 23 opinionated tools that serve as CEO, Designer, | Dev tools | TypeScript | 123,214 | +245 | Trending TypeScript · 245 stars today | heuristic seed 2026-07-20 |
Computer-use stack: agent desktops/VMs, cross-OS app driver, small models and Cua Bench | Agents | Rust | 29,853 | +7,248 this month | Gives agents macOS/Linux/Windows desktops via CLI, MCP or SDKs, plus Cua Bench to build tasks, evaluate agents and export trajectories. ⚠ "Teleport" moves signed-in app sessions into agent VMs; installs via curl | sh | catch-up · API-verified 2026-10-15 |
Self-hostable Go knowledge platform: RAG with citations, a tool-using agent, and a wiki | RAG / memory | Go | 34,186 | +225 | Swappable LLMs, vector DBs and storage; 29 model vendors, built-in MCP server, RBAC and Langfuse tracing; Docker Compose or K8s. ⚠ License NOASSERTION (MIT except listed third-party parts); browser-driving agent widens the attack surface | catch-up · API-verified 2026-10-15 |
Code review tool | Dev tools | Go | 10,977 | +162 | Hybrid architecture with LLM | Groq · llama-3.3-70b-versatile 2026-07-23 |
Docker CLI plugin to define, run and share multi-agent setups from declarative YAML | Agents | Go | 4,514 | +1,116 this week | Provider-agnostic agent runtime from Docker: MCP tools, built-in RAG, OCI-registry sharing; ships in Docker Desktop 4.63+. ⚠ Sends anonymous usage telemetry (stated in the README) | catch-up · API-verified 2026-10-15 |
Local runtime and REST API for running open-weight models, built on llama.cpp | Serving | Go | 183,287 | +1,019 this week | A common local backend routers sit in front of; one command wires it into Claude Code, Codex, OpenCode and other agents. | catch-up · API-verified 2026-10-15 |
Vectorless RAG: builds a tree index per document and has an LLM reason its way through it | RAG / memory | Python | 39,137 | +3,618 this month | An alternative to embedding-similarity retrieval for long PDFs; local SDK mode runs on your own LLM key, with an open benchmark repo. ⚠ OCR, MCP server and corpus-scale search are Cloud-only (paid key); benchmarks vendor-published | catch-up · API-verified 2026-10-15 |
A free, open source, and extensible speech-to-text application that works completely offli | Other | Rust | 27,034 | +107 | Trending Rust · 107 stars today | heuristic seed 2026-07-20 |
Scans agent skills and MCP tools for prompt injection, exfiltration and supply-chain risk | Dev tools | Python | 20,282 | +3,089 this month | 71 patterns, static plus optional LLM pass, SARIF/JSON output and a 0-100 risk score for gating Claude Code/Codex skill installs ⚠ Optional LLM pass sends skill content to a model provider; use --no-llm to stay static | catch-up · API-verified 2026-10-15 |
Python SDK and self-hosted proxy that calls 100+ LLM providers in the OpenAI format | Serving | Python | 61,324 | +2,719 this month | Comparison point for any router: virtual keys, spend tracking, guardrails, load balancing and an admin dashboard. ⚠ GitHub reports license NOASSERTION; enterprise features are a paid tier | catch-up · API-verified 2026-10-15 |
better agentic engineering | Agents | TypeScript | 3,383 | +79 | Trending TypeScript · 79 stars today | heuristic seed 2026-07-20 |
🚀 The fast, Pythonic way to build MCP servers and clients. | Agents | Python | 26,522 | +77 | Trending Python · 77 stars today | heuristic seed 2026-07-20 |
Rust inference engine for local LLMs on Apple Metal/CPU; Swift, Python and TS bindings | Infra | Rust | 2,349 | +503 this week | Has an OpenAI-compatible server and a bench CLI; a candidate local backend on Apple silicon, Linux targets still in progress. ⚠ Metal and CPU backends only today; Linux/Windows targets in progress | catch-up · API-verified 2026-10-15 |
NVIDIA library for quantization, pruning, distillation, sparsity and speculative decoding | Infra | Python | 5,326 | +1,531 this month | Exports quantized checkpoints (FP8, NVFP4) for vLLM, SGLang and TensorRT-LLM, the levers that set throughput and cost per token. ⚠ NVIDIA-centric formats (NVFP4) and integrations; figures vendor-reported | catch-up · API-verified 2026-10-15 |
Retrieval-augmented generation | RAG / memory | Python | 38,013 | +33 | Simple and fast RAG | Groq · llama-3.3-70b-versatile 2026-07-23 |
ML and LLM observability | Dev tools | Jupyter Notebook | 7,739 | +12 | Evaluate and monitor AI systems | Groq · llama-3.3-70b-versatile 2026-07-23 |
Kubernetes inference router: Envoy ext-proc endpoint picker with KV-cache-aware scheduling | Serving | Go | 407 | +26 this week | Shows in-cluster inference-aware routing: load and prefix-cache scoring, request priority, prefill/decode disaggregation. ⚠ Needs Kubernetes plus Envoy ext-proc; not a standalone gateway | catch-up · API-verified 2026-10-15 |