GitHub Trending Weekly Digest — June 16-20, 2026

The week of June 16-20 brought us an exciting mix of AI infrastructure, developer tooling, and cloud-native platforms. From AI-powered code intelligence to real-time geopolitical monitoring, this week's trending repos showcase the cutting edge of open source innovation.

🔥 Trending Titans (4+ Days on the Charts)

None this week — but several repos showed impressive 2-3 day runs.

💎 Multi-Day Champions (2-3 Days)

Agent-Reach

🔗 github.com/Panniantong/Agent-Reach
Days on trending: 2 (Jun 16, 18)

What it does:
One-line install to give AI agents full internet access — Twitter, Reddit, YouTube, GitHub, Xiaohongshu, Bilibili — all without API fees.

Why it matters:
AI agents can write code and manage projects, but accessing real-world internet content has been a nightmare. Every platform requires separate API setup (usually paid), anti-scraping workarounds, and login state management. Agent-Reach collapses all of this into a single CLI tool. It uses a "primary + fallback" multi-backend routing system — when one scraper breaks (like yt-dlp getting blocked by Bilibili in June 2026), it automatically switches to the next (bili-cli). Works with Claude Code, OpenClaw, Cursor, Windsurf, and any agent that runs shell commands.

Tech:
Python CLI, multi-platform routing (Jina Reader, yt-dlp, bili-cli, twitter-cli, OpenCLI, rdt-cli, xiaohongshu-mcp, gh CLI, Exa semantic search), MCP (Model Context Protocol) integration, local cookie storage.

codebase-memory-mcp

🔗 github.com/DeusData/codebase-memory-mcp
Days on trending: 2 (Jun 18, 20)

What it does:
Lightning-fast code intelligence MCP server that indexes your codebase into a persistent knowledge graph. Average projects finish in milliseconds; Linux kernel (28M lines) in 3 minutes.

Why it matters:
AI coding assistants traditionally use "file-by-file search + text grep" to understand code, causing massive token waste. Five structured queries with traditional methods burn ~412,000 tokens reading entire files; the knowledge graph approach uses just ~3,400 tokens (99.2% reduction). Queries return in <1ms for graph traversal, <10ms for name search and call path tracing. Single binary, zero dependencies, auto-configures for 11 programming agents including Claude Code, Codex, Zed, VS Code, OpenClaw, and Kiro.

Tech:
C (core engine), tree-sitter (158 language AST parsing), Hybrid LSP (semantic type resolution for Python/TS/PHP/C#/Go/C/C++/Java/Kotlin/Rust), embedded SQLite (knowledge graph storage), Cypher-like query language, Nomic nomic-embed-code embeddings (40K tokens, 768-dim int8), 3D graph visualization UI (Three.js).

Published research: Codebase-Memory: Tree-Sitter-Based Knowledge Graphs for LLM Code Exploration via MCP (arXiv:2603.27277) — 83% answer quality, 10× fewer tokens, 2.1× fewer tool calls across 31 real-world projects.

freeCodeCamp/freeCodeCamp

🔗 github.com/freeCodeCamp/freeCodeCamp
Days on trending: 2 (Jun 17, 19)

What it does:
Free, open-source platform teaching full-stack development and machine learning. Over 100,000 people have landed their first developer job through the curriculum.

Why it matters:
Coding bootcamps cost thousands of dollars, creating barriers for career-switchers. freeCodeCamp offers a complete, structured path from responsive web design to Python, databases, and backend development — all free, with interactive challenges, project-based learning, and industry-recognized certifications. Community support includes forums (help within hours) and YouTube courses.

Tech:
JavaScript/TypeScript, Node.js, React, PostgreSQL, Elixir. Curriculum covers HTML/CSS, JavaScript, React, Python, relational databases, backend APIs, plus language certifications (A2/B1 English, A1 Spanish/Chinese aligned to international standards). Also includes supplementary resources from The Odin Project, Coding Interview Prep, Project Euler, and Microsoft C# fundamentals.

iroh

🔗 github.com/n0-computer/iroh
Days on trending: 2 (Jun 18, 19)

What it does:
"IP addresses change. Public keys don't." — Rust-based modular P2P networking stack that uses public keys for dialing instead of IPs.

Why it matters:
Traditional P2P networks rely on IP addresses and ports, which change constantly due to NAT, dynamic allocation, and mobile network switching. Setting up NAT traversal, relay servers, and encrypted connections manually is brutally complex, forcing most apps to rely on centralized servers. Iroh uses public key addressing — connect via public key, and the system automatically finds the fastest route to the device, even if the IP changes. Auto NAT hole-punching with fallback to public relay servers. Performance continuously benchmarked at perf.iroh.computer.

Tech:
Rust, QUIC protocol (authenticated encryption, multiplexing, datagram delivery, no head-of-line blocking), Ed25519 public keys (identity + connection auth), NAT traversal (UDP hole-punching, STUN/TURN), modular protocols: iroh-blobs (BLAKE3 content-addressed transfer), iroh-gossip (pub-sub overlay network), iroh-docs (eventually-consistent key-value store).

iptv-org/iptv

🔗 github.com/iptv-org/iptv
Days on trending: 2 (Jun 16, 17)

What it does:
Global collection of publicly available IPTV (Internet Protocol TV) channels.

Why it matters:
Cable/satellite TV is expensive and region-locked. This project aggregates publicly available livestream URLs into M3U playlists. Users paste the playlist link into any video player (VLC, Kodi, etc.) to watch TV channels from around the world for free. Includes EPG (Electronic Program Guide) and API support.

Tech:
TypeScript, M3U playlist format, pure link aggregation (no video storage), data maintained by iptv-org/database.

Meshery

🔗 github.com/meshery/meshery
Days on trending: 2 (Jun 16, 18)

What it does:
Cloud-native management platform — visual GitOps self-service platform for managing all Kubernetes-based infrastructure and applications (multi-cloud, multi-cluster). Supports 380+ cloud-native integrations.

Why it matters:
Enterprises typically run hundreds of microservices across multiple clouds (AWS, GCP, Azure) and Kubernetes clusters (dev/test/prod). Traditional management involves: YAML hell (thousands of error-prone config lines), fragmented tooling (every cloud/service mesh/CI-CD has separate consoles), config drift (inconsistent multi-cluster settings), and steep learning curves (every tool has unique APIs). Meshery provides a single control plane for managing multi-cloud/multi-cluster Kubernetes, visual designer (drag-drop architecture without writing YAML), GitOps collaboration (IaC with version control and PR reviews), and curated design templates (best-practice configs for service meshes, monitoring stacks).

Tech:
React + TypeScript + Next.js (frontend), Go (backend, Kubernetes API interaction, performance testing), extension system (gRPC adapters for plugin integrations, hot-reload React bundles + Golang plugins, GraphQL + REST API, NATS message subscriptions), performance testing (Fortio load generation + Prometheus metrics + Grafana visualization), policy engine (OPA - Open Policy Agent), GitOps support (GitHub Actions, GitLab CI, Flux, ArgoCD).

CNCF incubating project with active community.

teslamate-org/teslamate

🔗 github.com/teslamate-org/teslamate
Days on trending: 2 (Jun 16, 17)

What it does:
Self-hosted Tesla vehicle data logger providing high-precision drive data capture and visualization.

Why it matters:
Tesla owners want deep insights into vehicle data (battery health, energy consumption, charging statistics, drive routes), but the official app offers limited functionality. TeslaMate enables local deployment with zero vampire drain, MQTT integration with Home Assistant, custom geofencing, and comprehensive analytics.

Tech:
Elixir, PostgreSQL, Grafana (visualization), MQTT, Docker. AGPLv3 license to prevent closed-source commercial forks.

TimesFM

🔗 github.com/google-research/timesfm
Days on trending: 2 (Jun 19, 20)

What it does:
Google Research's pretrained Time-Series Foundation Model for forecasting.

Why it matters:
Traditional time-series forecasting requires training specialized models for each scenario — high cost, low generalization. TimesFM is a foundation model that works out-of-the-box for diverse forecasting tasks without retraining. TimesFM 2.5 supports up to 16k context length (massive jump from 2048), handling longer historical sequences. Includes 30M-parameter quantile head for continuous quantile forecasting (10th-90th percentiles), covering horizons up to 1k. Already integrated into Google products (BigQuery ML, Google Sheets, Vertex Model Garden) — forecast with SQL queries.

Tech:
Decoder-only Transformer, 200M parameters (down from 500M), PyTorch + Flax (JAX) implementations, CPU/GPU/TPU/Apple Silicon support, HuggingFace Transformers + PEFT (LoRA) fine-tuning examples, XReg for external feature inputs.

Based on paper: A decoder-only foundation model for time-series forecasting (ICML 2024). Checkpoints: Hugging Face Collection. Community contributor @borealBytes built Agent integration for programming assistants.

🌟 Single-Day Stars

aishwaryanr/awesome-generative-ai-guide

🔗 github.com/aishwaryanr/awesome-generative-ai-guide
Spotted: Jun 20

One-stop resource hub for Generative AI: research updates, interview prep, Jupyter notebooks, and structured curriculum.

What it offers:
Structured learning roadmaps (5-day LLM fundamentals, 3-day RAG, 5-day Agents), 90+ free GenAI courses, 60 GenAI interview questions with answers, monthly best papers list, ICLR 2024 paper summaries.

Free certified courses by Aishwarya Naresh Reganti & Kiriti Badam: Applied LLMs Mastery 2024 (10-week complete course), Generative AI Genius 2024, AI Evals for Everyone, OpenClaw Mastery for Everyone.

1000+ students participated in Applied LLMs Mastery 2024. Regular updates with best monthly papers and conference summaries.

chatwoot/chatwoot

🔗 github.com/chatwoot/chatwoot
Spotted: Jun 16

Open-source omnichannel customer engagement platform — the Intercom/Zendesk/Salesforce Service Cloud alternative.

Consolidates all customer conversations into one inbox: website live chat, email, Facebook, Instagram, Twitter, WhatsApp, Telegram, Line, SMS. Built-in AI assistant "Captain" handles common questions. Includes help center portal, team collaboration, automation workflows, CRM, and reporting.

Tech: Ruby on Rails, PostgreSQL, Redis, Sidekiq, WebSocket, Docker, Kubernetes.

GLM-5

🔗 github.com/zai-org/GLM-5
Spotted: Jun 19

Zhipu AI's latest flagship model GLM-5.2, designed for long-horizon tasks and agentic engineering.

Traditional models plateau early in long tasks. GLM-5.2 supports stable 1M token context, enabling continuous optimization across hundreds of rounds and thousands of tool calls. Achieves open-source SOTA on real-world coding tasks (SWE-Bench Pro, Terminal-Bench 2.1). Ranks #1 among open-source models on Vending Bench 2 (simulates one-year vending machine operation).

Tech: 744B parameters (40B active, MoE architecture), pretrained on 28.5T tokens, integrates DeepSeek Sparse Attention (DSA) for lower deployment cost, IndexShare architecture (every 4 sparse attention layers share indexer, 2.9× FLOPs reduction at 1M context), improved MTP layer for speculative decoding (20% longer acceptance length), supports multiple reasoning budgets (reasoning_effort: max/high), deployment via SGLang, vLLM, Transformers, KTransformers, Ascend NPU support.

Performance highlights: Terminal-Bench 2.1: 81.0 (close to Claude Opus 4.8's 85.0), SWE-bench Pro: 62.1, Vending Bench 2: $4,432 final balance (open-source #1, close to Claude Opus 4.5).

palmier-io/palmier-pro

🔗 github.com/palmier-io/palmier-pro
Spotted: Jun 20

macOS-native video editor for AI, built from scratch in Swift. Premiere Pro for the AI era with AI generation + MCP protocol integration for programming assistants.

Directly invoke Seedance/Kling/Nano Banana Pro from the timeline to generate video and images. Exposes MCP server via HTTP (127.0.0.1:19789/mcp), letting Claude Code/Codex/Cursor control timeline and editing logic. macOS 26 (Tahoe) + Apple Silicon required.

Open-source: Video editor core + MCP server + Agent chat interface (GPLv3). Closed-source: Generative AI processing modules. Free: Editor + MCP server (no login required). Paid: Generative AI features require subscription.

Tech: Swift (macOS native), built-from-scratch video editing engine (non-framework-based), AI model integrations (Seedance, Kling, Nano Banana Pro), MCP server (HTTP transport at http://127.0.0.1:19789/mcp), Claude Desktop Extension support (one-click install via mcpb).

Community: Discord | Twitter @Palmier_io | YC Company Page

puppeteer/puppeteer

🔗 github.com/puppeteer/puppeteer
Spotted: Jun 17

Google's official Node.js library controlling Chrome/Firefox via DevTools Protocol or WebDriver BiDi.

Automates browser operations (UI testing, screenshots, scraping, form autofill) with stable high-level API. Supports headless mode, making it the go-to tool for E2E tests and automation scripts.

Tech: TypeScript/JavaScript, Chrome DevTools Protocol, WebDriver BiDi. Integrates with MCP server (chrome-devtools-mcp).

superpowers

🔗 github.com/obra/superpowers
Spotted: Jun 19

Skills framework and software development methodology for AI coding agents, supporting mainstream assistants (Claude Code, Cursor, Codex, etc.).

AI assistants often skip design phases and jump straight to code, causing drift. Superpowers provides a complete methodology: requirement clarification → design review → implementation plan → subagent-driven development → code review.

Skill system: brainstorming (design discussion), writing-plans (task decomposition), subagent-driven-development (concurrent subagent execution). Mandatory workflows: test-driven-development (TDD), systematic-debugging, using-git-worktrees (parallel branching). Supported platforms: Claude Code, Antigravity, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, Pi. Includes evaluation framework (superpowers-evals).

Core principles: Test-first (RED-GREEN-REFACTOR), systematic over ad-hoc, simplify complexity, evidence over claims.

swc-project/swc

🔗 github.com/swc-project/swc
Spotted: Jun 17

Rust-based super-fast TypeScript/JavaScript compiler, providing libraries for both Rust and JavaScript.

Traditional JS toolchains (like Babel) are slow; SWC rewrites the core in Rust for extreme performance — 20× faster than Babel. Provides code transformation, minification, and bundling.

Tech: Rust, JavaScript/TypeScript, MSRV 1.73+, supports Node v10+ usage and Node v20+ development.

Universal Android Debloater Next Generation

🔗 github.com/Universal-Debloater-Alliance/universal-android-debloater-next-generation
Spotted: Jun 18

Cross-platform GUI tool removing bloatware from non-rooted Android devices via ADB, improving privacy, security, and battery life.

Android devices (especially vendor-customized systems) come with tons of uninstallable system apps (bloatware) that waste storage/memory, collect privacy data, drain battery, and slow performance. Traditional root methods are risky and void warranties; regular users struggle to safely remove these apps.

How it works: Root-free debloating via ADB (Android Debug Bridge), no bootloader unlock or root required. Community-driven Universal Debloat List annotates safety levels for each system app (safe to remove / cautious / not recommended). Cross-platform GUI (Rust + Iced framework) for Windows, macOS, Linux. Reversible operations — apps are "disabled" not physically deleted, can restore anytime.

Tech: Rust (high-performance, memory-safe), Iced GUI library, ADB (pm uninstall -k --user 0 command to disable system apps), data source: uad_lists.json with thousands of system app safety classifications.

Use cases: Remove vendor preloaded software (Xiaomi, Samsung, Huawei ad apps), extend old device battery and performance (remove background autostart services), improve privacy (remove data collection components, vendor analytics SDKs).

Note: Read usage guide and app safety descriptions before operating. Incorrect removal of critical system components may cause device malfunction (but can be fixed via restore).

worldmonitor

🔗 github.com/koala73/worldmonitor
Spotted: Jun 20

Real-time global intelligence dashboard: AI-driven news aggregation + geopolitical monitoring + infrastructure tracking in a unified situational awareness interface.

Integrates 500+ curated news sources (15 categories), AI-synthesized briefings. Country Instability Index (CII v8) real-time assessment of 31 tier-1 countries' stress scores, correlating military/economic/disaster/escalation signals. Covers 29 stock exchanges + commodities + crypto, 7-signal market composite indicators. Dual map engines (3D globe via globe.gl + WebGL flat map via deck.gl), 56 map layer types.

Tech: Vanilla TypeScript + Vite, globe.gl (Three.js) + deck.gl (MapLibre GL), Tauri 2 desktop app (Rust framework) + Node.js sidecar, AI/ML via Ollama (local AI, zero API key) / Groq / OpenRouter, Transformers.js (browser-side inference), Protocol Buffers (276 protos, 34 services), sebuf HTTP annotations, deployment on Vercel Edge Functions (60+ functions), Railway relay, PWA, self-hosted (Docker/static), Redis cache (Upstash) + 3-tier cache + CDN + Service Worker.

Features: 6 site variants (world / tech / finance / commodity / happy / energy) from single codebase, 24 language support (native source + RTL), desktop app (single Tauri binary, in-app variant switching for Windows / macOS Intel & ARM / Linux AppImage), local-first (runs via Ollama, no cloud required), data coverage: 65+ external providers, flight data from Wingbits.

Live demos: Main | Tech | Finance | Commodity | Energy

License: AGPL-3.0-only (commercial use allowed, must comply with open-source obligations).

📊 Week in Themes

🤖 AI Infrastructure Week

Four AI-first tools dominated: Agent-Reach (internet access for AI), codebase-memory-mcp (code intelligence), TimesFM (time-series foundation model), GLM-5.2 (long-horizon agentic tasks). Pattern: Foundation models + agent tooling as critical infrastructure.

🎥 Content & Media Revolution

Palmier Pro brings AI generation directly into video editing timelines with MCP protocol integration. IPTV aggregates global livestreams. Theme: Democratizing professional media tools.

🔒 Privacy & Local-First Computing

Universal Android Debloater removes tracking bloatware without root. WorldMonitor runs fully local with Ollama. TeslaMate keeps car data on your own server. Pattern: User sovereignty over personal data.

☁️ Cloud-Native Management

Meshery (380+ integrations), iroh (P2P networking), codebase-memory-mcp (local code graphs). Theme: Complex infrastructure made manageable through abstraction layers.

📚 Learning Ecosystems

freeCodeCamp (100K+ job placements), awesome-generative-ai-guide (structured GenAI paths). Pattern: Community-driven education at scale.

Key Takeaway

This week's trending repos reflect a maturing AI ecosystem moving beyond model capabilities toward integration and infrastructure. The standout pattern: tools that reduce friction between AI and existing workflows (Agent-Reach's one-line internet access, codebase-memory-mcp's 99% token savings, Palmier Pro's MCP integration). As AI agents become more capable, the bottleneck shifts from intelligence to connectivity — and this week's projects are solving that.

Compiled by Tommy Zhang | June 21, 2026