⭐️ We built Prompt Management for teams shipping AI agents.
• Cloud-hosted prompts
• Version history & rollbacks
• Environment labels (dev, staging, production)
• A/B testing
• Usage analytics
• Audit trail
Treat prompts like code, without keeping them buried in your application source.
Built a Slack AI agent example with VoltAgent Chat SDK.
Thread-aware conversations, tool calling, interactive actions, Redis persistence, and Next.js webhooks included.
npm create voltagent-app@latest – –example with-chat-sdk
38 YouTube videos, hundreds of shorts, and tweets from people we’ve never talked to, all in the first month of this side project.
Feels like we're starting to smell real hype.
We’re launching designmd.sh — a public DESIGN.md registry.
"npx designmd.sh add <owner/repo>"
@stitchbygoogle team introduced DESIGN.md as a way to describe design systems in a format agents can read.
Through awesome-design-md, we were one of the first teams to publish real-world DESIGN.md files and bring the idea into the open source community.
Now we’re taking the next step: the first public registry layer for DESIGN.md files.
Anyone can list and showcase their own public DESIGN.md files hosted on GitHub.
designmd.sh
New update — @voltagent/core@2.7.1 👀
Agents can now adapt per request using headers.
→ switch model
→ change instructions
→ inject tools
All dynamically.
An awesome repo for DESIGN.md files already existed.
Those files speak Claude Design's native tongue, so we gave them a proper home: Awesome Claude Design 68 files with installation guides.
One-shot prompts, rubrics, animation recipes, and export recipes for Claude Design are next.
✍️ github.com/VoltAgent/awesome…
⭐️ 47.000 GitHub stars ⭐️
Less than 1 year.
$0 marketing spend.
We didn't write a book.
We shipped what we needed and shared it.
If you're waiting for a book…
don't.
We're busy building ⚡️
Protecting your agent endpoints should not be complicated.
VoltAgent’s authentication recipe shows how to use authNext to separate:
• public routes
• console routes
• user routes
with JWT auth, console access keys, and custom public endpoints.
Clean setup, practical defaults, and easy to extend.
Docs: voltagent.dev/recipes-and-gu…
DESIGN. md concept introduced by @stitchbygoogle .
Our repo awesome-design-md brought it to the community, 35K stars in one week, hundreds of organic shares.
Now it has a website: getdesign.md/
Browse design systems as DESIGN. md inspired by real websites.
Preview instantly. Install with one command.
One markdown file tells your AI agent exactly how to build matching UI.
voltagent/core 2.7.0 is out 🔋
→ prepareStep — control which tools are available at each step, right from agent creation. no more repeating callbacks on every call.
→ serverless fix — tools crashing on Cloudflare Workers / Vercel Edge because cleanup ran too early? fixed. waitUntil now defers properly.
The DESIGN.md introduced by @stitchbygoogle : a plain markdown file that tells AI agents exactly how your UI should look.
Think of it as README.md/AGENTS.md for design.
Colors, fonts, spacing, components, all in one file any LLM can read.
We're building an awesome repo of DESIGN.md files inspired by popular websites. Not official design systems, but close enough for AI agents to generate matching UI.
Starting with developer-focused sites (Vercel, Linear, Stripe, Supabase, Cursor...) and expanding to every site with a beautiful design system.
Copy one into your project. Your AI agent does the rest.
here are the top 5 fastest growing github repos this week:
1. HKUDS/ClawTeam (2.7K stars)
clawteam: agent swarm intelligence (one command → full automation)
2. VoltAgent/awesome-codex-subagents (2.0K stars)
a collection of 130 specialized codex subagents covering a wide range of develo
3. MiniMax-AI/skills (1.3K stars)
no description
4. math-inc/OpenGauss (1.0K stars)
no description
5. lcoutodemos/clui-cc (1.0K stars)
clui cc — command line user interface for claude code
Working with @voltagent_dev has been amazing.
Big shout out to @omerfarukaplak who's been pushing updates non-stop and responds to feedback incredibly fast.
If you're building AI agents, give it a try.
VoltOps now supports MCP.
You can now plug AI agents into VoltOps via MCP and let them inspect real traces and logs.
Instead of guessing from code, they can see what actually happened.
Ask things like “why did this run fail?” or “what’s slow today?” and it will dig through traces and logs for you.