The workbench

Trending repos.

What builders are starring this week, agents, sandboxes, MCP tooling. Tap any to open it on GitHub.

GitHub trending

23 repos · updated weekly
01stablyai/orcaAn ADE to run a fleet of coding agents in isolated git worktrees.note Breaking out this window. · Anyone running Claude Code / Codex / OpenCode side-by-side+5.7k/wk02iOfficeAI/OfficeCLIA CLI for agents to edit Office documents with no Office install.note Breaking out this window. · Headless document automation+5.3k/wk03ogulcancelik/herdrTerminal multiplexer for many AI agents, with detachable sessions.Devs juggling long-running agents~17.4k (+2.5k/wk)04TencentCloud/CubeSandboxSub-60ms hardware-isolated sandbox for agent-generated code.Builders running model-generated code~10.4k (+1.1k/wk)05calesthio/OpenMontageTurns your coding assistant into a video production studio (12 pipelines, 52 tools, 500+ skills).note Hit #1 Repo of the Day. AGPLv3. · Content teams + indie builders wanting programmatic video from prompts~+3.7k in a day06bytedance/deer-flowByteDance's ground-up long-horizon agent (persistent memory, sub-agents) for multi-hour tasks.note v2.0.0 shipped Jun 25. · Builders wanting a batteries-included long-horizon agent harness07tirth8205/code-review-graphLocal Tree-sitter code graph as an MCP server; ~82x median token reduction by feeding agents only blast-radius files.Teams burning tokens running agents on big monorepos08heygen-com/hyperframes"Write HTML, render video": deterministic HTML→MP4.note v0.7.48 on Jul 10. · Reproducible code-defined video in agent workflows09vercel-labs/agent-browserFast native Rust browser for agents; local + cloud browsers.A lightweight alternative to a heavy Playwright/Puppeteer stack10jamiepine/voiceboxLocal, on-device AI voice stack (voice I/O for agents).note Broke out in-window. · Privacy-conscious voice / local-first builders11tashfeenahmed/freellmapiOpenAI-compatible gateway to free providers (~161 models) behind one /v1 with routing/failover.note Check provider ToS. · Prototypers wanting broad model access12google-labs-code/design.mdGoogle Labs spec giving agents a persistent design-system understanding; WCAG linting, Tailwind export.Frontend teams wanting on-brand agent output13JCodesMore/ai-website-cloner-templateReverse-engineers a site into clean Next.js 16 / React 19.note Repo warns against impersonation use. · Devs/agencies doing rebuilds14NousResearch/Hermes-AgentSelf-improving agent with autonomous skill-learning, 300+ models, cron, and parallel subagents.note Flagged breaking-out, not new; v0.18.x shipped Jul 1-8 but predates the 14-day window. · Builders wanting an autonomous, self-improving agent harness15Graphify-Labs/graphifyTurns any folder of code, SQL, docs, or media into a queryable knowledge graph; 100% local tree-sitter parsing across 33 languages.note Star count exploding between fetches. · Anyone running an agentic coder on a big codebase~58-83k16can1357/oh-my-piRust terminal coding agent with hash-anchored edits, LSP, browser control, and subagents.Devs wanting a powerful open-source terminal agent~17.5k17microsoft/SkillOptTrains reusable natural-language skills for frozen LLM agents, reporting a 19 to 25 point lift.Teams building agent skills who want measurable optimization~2.4k18t8y2/dbx60+ databases in a 20MB Rust client with a built-in MCP server.Devs wanting one fast DB client that doubles as an agent data source~9.9k19bradautomates/claude-videoThe /watch command gives Claude video comprehension via frames plus transcript.Claude Code users needing video understanding~7.9k20ComposioHQ/awesome-claude-skills1,000+ production-ready Claude Skills plus authoring templates.note Curated list, moves fast. · Anyone building on Claude Skills~67.6k21777genius/agent-teams-aiDesktop app to orchestrate parallel agent teams on a kanban board.Builders experimenting with multi-agent orchestration~1.6k22traycerai/traycerDesktop multi-agent orchestration with shared context and mid-conversation model switching.note Early. · Early adopters wanting a multi-model coordination hub~50023simonw/llm-coding-agentA minimal, hackable coding agent (file edit, command exec, search, Python API) on the llm CLI.note Reference implementation; deliberately small. · People rolling their own agent instead of buying one

Get the ones that matter, weekly.

Stay on top of what shipped in AI. Weekly update.

Unsubscribe anytime
~/subscribe $