The Jan 2026 top TypeScript repos on GitHub are a strong signal for where the ecosystem is heading. https://github.com/trending/typescript?since=monthly
1. openclaw/openclaw
One-line description: Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
Summary:
- OpenClaw is a personal AI assistant you run on your own devices, answering on the channels you already use (WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Microsoft Teams, WebChat, and more).
- It includes a Gateway control plane, multi-channel inbox, voice wake/talk mode for macOS/iOS/Android, and a live Canvas you can control; the Pi agent runs in RPC mode with tool and block streaming.
- Install via
openclaw onboard --install-daemon; works with npm, pnpm, or bun and supports OAuth for Anthropic and OpenAI.
2. anomalyco/opencode
One-line description: The open source coding agent.
Summary:
- OpenCode is an open source AI coding agent similar in capability to Claude Code, with 100% open source code and no lock-in to a single provider (works with Claude, OpenAI, Google, or local models).
- It ships with two built-in agents: build (full access for development) and plan (read-only for analysis and exploration, with permission prompts before bash).
- Available as CLI (
curl -fsSL https://opencode.ai/install | bash) and as a desktop app (BETA) for macOS, Windows, and Linux.
3. remotion-dev/remotion
One-line description: Make videos programmatically with React.
Summary:
- Remotion is a framework for creating videos using React: you use components, composition, and the web stack (CSS, Canvas, SVG, WebGL) to define frames and timelines.
- You get variables, functions, APIs, and the npm ecosystem instead of a traditional video editor; Fast Refresh and the React model apply to video composition.
- Get started with
npx create-video@latest; note that Remotion has a special license and may require a company license in some cases.
4. Lissy93/web-check
One-line description: All-in-one OSINT tool for analysing any website.
Summary:
- Web-Check gives a single dashboard to inspect a site’s infrastructure and security: IP info, SSL chain, DNS records, cookies, headers, crawl rules, server location, redirect chain, open ports, traceroute, DNSSEC, performance, trackers, carbon footprint, and more.
- It helps you understand, optimize, and secure a website; optional API keys (e.g. Google, Shodan, WhoAPI) unlock extra checks and higher rate limits.
- Deploy via Netlify, Vercel, Docker (
docker run -p 3000:3000 lissy93/web-check), or from source; live demo at web-check.xyz.
5. bytedance/UI-TARS-desktop
One-line description: The Open-Source Multimodal AI Agent Stack: Connecting Cutting-Edge AI Models and Agent Infra.
Summary:
- The repo ships two things: Agent TARS (multimodal AI agent with CLI and Web UI, GUI + vision, MCP tools) and UI-TARS Desktop (native desktop app that controls your computer via natural language and the UI-TARS / Seed-1.5-VL models).
- Agent TARS can book flights, hotels, and draw charts using real browsers and MCP; UI-TARS Desktop supports local and remote computer/browser operators with screenshot recognition and precise mouse/keyboard control.
- Run with
npx @agent-tars/cli@latestor install the desktop app; supports Volcengine/Anthropic and others.
6. virattt/dexter
One-line description: An autonomous agent for deep financial research.
Summary:
- Dexter turns complex financial questions into step-by-step research plans, runs them with live market data, and refines answers with self-validation and loop/step limits.
- It uses intelligent task planning, autonomous tool execution, and access to income statements, balance sheets, and cash flows; optional Exa or Tavily keys improve web search.
- Built with Bun; run
bun startfor interactive mode; evals and LangSmith logging are included; tool calls are logged to.dexter/scratchpad/for debugging.
7. badlogic/pi-mono
One-line description: AI agent toolkit: coding agent CLI, unified LLM API, TUI & web UI libraries, Slack bot, vLLM pods.
Summary:
- Pi monorepo provides a unified multi-provider LLM API (pi-ai), agent runtime with tools (pi-agent-core), interactive coding agent CLI (pi-coding-agent), Slack bot (pi-mom), TUI library (pi-tui), web chat components (pi-web-ui), and vLLM pod management (pi-pods).
- It’s the foundation used by OpenClaw and similar projects; you get one stack for building and running AI coding agents and UIs.
- Develop with
npm install,npm run build,npm run check; run the coding agent from repo root with./pi-test.sh.
8. numman-ali/openskills
One-line description: Universal skills loader for AI coding agents — npm i -g openskills.
Summary:
- OpenSkills brings Anthropic’s skills system (SKILL.md) to every AI coding agent (Claude Code, Cursor, Windsurf, Aider, Codex, etc.) via a universal installer that syncs skills into AGENTS.md.
- You install from the Anthropic marketplace, any GitHub repo, or local path; agents load skills on demand with
npx openskills read <name>, keeping context small. - Use
npx openskills install anthropics/skillsandnpx openskills sync; supports--universalfor.agent/skills/in multi-agent setups.
9. EveryInc/compound-engineering-plugin
One-line description: Official Claude Code compound engineering plugin.
Summary:
- Compound Marketplace is a Claude Code plugin that implements compound engineering: Plan → Work → Review → Compound, so each unit of work makes the next easier via planning, review, and codified learnings.
- Commands include
/workflows:plan(feature → implementation plans),/workflows:work(execute with worktrees/tasks),/workflows:review(multi-agent code review), and/workflows:compound(document learnings). - Install in Claude Code with
/plugin marketplace add https://github.com/EveryInc/compound-engineering-pluginand/plugin install compound-engineering; experimental OpenCode/Codex conversion via Bun CLI is included.
10. cjpais/Handy
One-line description: A free, open source, and extensible speech-to-text application that works completely offline.
Summary:
- Handy is a cross-platform desktop app (Tauri: Rust + React/TypeScript) for privacy-focused speech-to-text: press a shortcut, speak, and have text pasted into any field without sending audio to the cloud.
- It uses local VAD (Silero), then Whisper or Parakeet models for transcription, with GPU acceleration when available; supports Windows, macOS, and Linux.
- Download from the releases page or handy.computer; designed to be simple and forkable for accessibility and extensibility.
11. danielmiessler/Personal_AI_Infrastructure
One-line description: Agentic AI Infrastructure for magnifying HUMAN capabilities.
Summary:
- PAI (Personal AI Infrastructure) is a goal-oriented, human-centric platform that turns Claude Code (or other agents) into a persistent assistant with memory, skills, and continuous learning rather than a stateless task runner.
- It uses TELOS (10 files for mission, goals, beliefs, etc.), a skill system (CODE → CLI → prompt → skill), memory and hooks, and installable Packs/Bundles; full release install copies a complete
.claude/in minutes. - Install via full release (recommended), Bundle + packs, or individual packs; supports voice (ElevenLabs), notifications (ntfy, Discord), and a terminal status line; goal is to activate people and make strong AI infrastructure widely accessible.
12. darrenhinde/OpenAgentsControl
One-line description: AI agent framework for plan-first development workflows with approval-based execution. Multi-language support (TypeScript, Python, Go, Rust) with automatic testing, code review, and validation built for OpenCode.
Summary:
- OpenAgentsControl (OAC) teaches agents your coding patterns via a context system (ContextScout); agents propose plans, you approve, then they implement with validation—aiming for production-ready code without heavy refactoring.
- It emphasizes approval gates (no execution without your OK), MVI token efficiency, editable agents (markdown), and team-shared context in
.opencode/context/project/committed to the repo. - Install with
curl -fsSL .../install.sh | bash -s developer; useopencode --agent OpenAgentor OpenCoder; works with any model and integrates with OpenCode.
13. lobehub/lobehub
One-line description: The ultimate space for work and life — to find, build, and collaborate with agent teammates that grow with you.
Summary:
- LobeHub is a work-and-life space for creating and collaborating with AI agents: agents as the unit of work, Agent Groups for teamwork, Pages/Schedule/Project/Workspace, and Personal Memory so agents learn from how you work.
- It offers 10,000+ skills/MCP plugins, multi-model and local LLM support (e.g. Ollama), Chain of Thought, branching conversations, artifacts, file/knowledge base, TTS/STT, text-to-image, and PWA/desktop.
- Self-host on Vercel, Zeabur, Sealos, Alibaba Cloud, or Docker; supports Better Auth and optional server-side DB; ecosystem includes @lobehub/ui, icons, TTS, and lint packages.
14. ChromeDevTools/chrome-devtools-mcp
One-line description: Chrome DevTools for coding agents.
Summary:
- chrome-devtools-mcp is an MCP server that gives AI coding assistants (Gemini, Claude, Cursor, Copilot, etc.) control over a live Chrome instance: automation, performance traces, network inspection, screenshots, and console messages.
- It uses Puppeteer for reliable automation and Chrome DevTools for performance insights; you can connect to an existing Chrome (e.g.
--browser-url=http://127.0.0.1:9222) or let it launch its own. - Add to any MCP client with
npx -y chrome-devtools-mcp@latest; supports headless, viewport, proxy, and category toggles; note that browser content is exposed to the MCP client.
15. nocodb/nocodb
One-line description: A Free & Self-hostable Airtable Alternative.
Summary:
- NocoDB turns MySQL, PostgreSQL, SQL Server, SQLite, and MariaDB into a smart spreadsheet UI: grids, gallery, form, Kanban, calendar, sharing, roles, and rich cell types (lookup, rollup, formula, attachment, etc.).
- It includes an App Store for workflow automations (chat, email, storage), REST APIs, and an SDK; you keep your data in your own DB and avoid vendor lock-in.
- Run with Docker (SQLite or PG), auto-upstall script, or binaries; mission is to be the most powerful no-code database interface for every internet business.
16. modelcontextprotocol/ext-apps
One-line description: Official repo for spec & SDK of MCP Apps protocol — standard for UIs embedded in AI chatbots, served by MCP servers.
Summary:
- MCP Apps extend MCP so tools can declare interactive UI resources (
ui://); the host fetches and renders them in a sandboxed iframe, with bidirectional communication between UI and host. - The repo provides the spec (SEP-1865), SDK for app developers (
@modelcontextprotocol/ext-appsand React hooks), and SDK for hosts (app-bridge); inspired by MCP-UI and OpenAI Apps SDK. - Install with
npm install -S @modelcontextprotocol/ext-apps; examples include map, Three.js, Sheet Music, budget allocator, PDF, and more; optional Claude Code plugin for building MCP Apps.
17. thedotmack/claude-mem
One-line description: A Claude Code plugin that automatically captures everything Claude does during your coding sessions, compresses it with AI, and injects relevant context back into future sessions.
Summary:
- Claude-Mem keeps context across sessions by capturing tool-use observations, generating semantic summaries, and storing them in SQLite (with optional Chroma for hybrid search); future sessions get relevant context automatically.
- It uses lifecycle hooks (SessionStart, UserPromptSubmit, PostToolUse, Stop, SessionEnd), a worker on port 37777 with web viewer and search API, and a mem-search skill plus MCP tools (search, timeline, get_observations) with progressive disclosure for token efficiency.
- Install via
/plugin marketplace add thedotmack/claude-memand/plugin install claude-mem; supports\<private>tags, configurable context injection, and beta features like Endless Mode; requires Node 18+, Claude Code, Bun, and optionally uv for vector search.
Source: GitHub trending TypeScript (monthly). Summaries based on project READMEs for January 2026.