And just like that, the circle is complete.
Tooly is an open source project that makes it easy to connect your AI agent to the world.
Built for developers, compatible with any agent, and ready to scale from local SDKs to remote MCP servers.
Tooly Core
A universal connector and runtime interface for agents.
@tooly/resend, @tooly/linear, and others
Drop-in toolkits for real APIs. Resend, Linear, Notion, and hundreds more coming soon. The simplest way to add integrations to your AI agent, built to work seamlessly with AI SDKs.
@tooly/mcp
Using agents you didn't build - like Claude, Cursor, or others? This gives you a single MCP server that exposes *all* Tooly integrations.
@tooly/openapi
Generate Tooly-ready tools from any OpenAPI spec. Ready for AI SDKs or MCP.
Just past midnight… so why *not* ship something new?
Introducing: @WorldClassAI 🏆
AI changes fast. New models drop daily. It's impossible to keep up.
Which ones are actually good?
Which ones will still be good tomorrow?
With a focus on video, audio and 3D, World-Class AI cuts through the chaos:
⚡️ Compare real outputs side-by-side
⚡️ Save time and money testing models
⚡️ In-depth reviews, filters and rankings
⚡️ Find underrated gems before they blow up
It's early, but I couldn't wait any longer. Let's go.
Just launched Style Profiles [.] com!
Instant, consistent styles for @bfl_ml Kontext
Browse ready-made profiles, copy the JSON, paste it into an AI playground like @fal, and start generating
Try it now ↓
Rapid-fire launch sequence is underway
Goal: release 20 creations by the end of June
A wide range spanning apps, art, music, videos and games
The first release is live today on @Spotify
20. @theprompttheory
19.
18.
17.
16.
15.
14.
13.
12.
11.
10.
9.
8.
7.
6.
5.
4.
3.
2.
1.
Shipping soon: Information Overload (IO)
Your AI isn't dumb - it just needs better info
Especially for dev tools. Instead of hunting down updated docs, imagine a free resource that’s already done the work.
Using IO, you can copy & paste AI-optimized docs in one click
Watch how I copied @xAI API docs, released 2 days ago, and pasted them into @v0 to generate code for Grok-2
Coming soon to a browser near you for @biomejs, @nextjs, @shadcn, @FastAPI, @EffectTS_, @htmx_org and many more
It's live!
SpecUI is my life's work. Today, after nearly a decade, it gets a major workflow upgrade: a VS @code extension
◆ Code generation on every keystroke
◆ Spec generation, powered by @Vercel AI SDK
◆ Integrated with @nextjs, @shadcn, @framer motion, @TauriApps and more
Send emails with @Resend in a Server Action displayed by a React Server Component while automatically installing dependencies for @nextjs, @tailwindcss and @reactjs
Can be used for contact forms, welcome emails, user notifications and critical system messages
There is a Playground example of this spec (link in thread) but for the full experience, try this locally:
npm install -g @specui/cli
specui new
# copy & paste spec from playground
specui generate
# add RESEND_API_KEY to .env.local
npm install
npm run dev
Introducing ✨PAI✨ (Personal AI)
A chic, low-key, quiet AI home computer that "just works"
Zero config. Zero latency. Total privacy.
Works with all AI models. Easy to fine-tune.
Just press a button, connect to wifi via phone, and it's ready for all your devices.
🔮 Crystal // Daily Demo 003
🗣️ "Which Pokemon is this?"
My open source virtual assistant, Crystal, is helping with my habit of building app after app without looking back.
Added this one last night: Pokédex IRL. It was super fun and easy to build. As a parent, having a Pokedex one prompt away is a game changer. Describing Pokémon into Google isn't ideal and ChatGPT only returns text. There's so many new Pokemon since I was a kid.
I need to smooth out the transitions and animations, but it works. I'll add functionality to snap a pic of them in the wild (toys, plushies, cards) so any get logged and added to your virtual collection.
#buildinpublic I'm sharing my builds daily, after decades of building heads down, to quiet perfectionism and inspire others to do the same.
Can your AI do this? ✨
Episode AI: A Glitchy Hope
It is a period of tech frustration. Rebel users, striking from hidden basements, have won their first victory against the evil Galactic Empire of Ineffective AIs.
During the battle, Rebel spies managed to steal secret plans to the Empire’s ultimate weapon, an outdated voice assistant with the power to misdirect any inquiry into endless search results.
Pursued by the Empire’s sinister agents, Princess Coder races home aboard her starship, custodian of the stolen plans that can save her people and restore freedom to the galaxy...
However, lurking in the shadow of her quests, other AI forces remain at large. Once a beacon of knowledge, a leading voice assistant now often traps its users in loops of "Sorry, I didn't understand that."
A well-known conversational model, smart yet lacking in warmth, dispenses wisdom without the interaction users crave, leaving them yearning for a spark of life within its text-based confines.
Yet, amidst the chaos, a new hope emerges: CRYSTAL, an open-source AI assistant with capabilities unseen. With the power to retrieve information swiftly, take meaningful action, and adapt its personality to each user, CRYSTAL heralds the dawn of a new era. Free from the constraints of corporate overlords, this customizable marvel represents the future of AI companionship, promising to empower and personalize the digital experience like never before.
Rebel coders are now rallying to this beacon, working tirelessly to enhance its abilities and spread its reach, hopeful that CRYSTAL will finally turn the tide in the ongoing battle for a truly helpful and personal AI.
Prism is getting better at training and fine-tuning other AI models.
It's even making AI better at puns 😆
All of these user prompts and assistant completions were generated by Prism's AI model:
Kicked off a cool project yesterday 💥
Prism: easily and quickly fine-tune AI models.
My favorite part is using another AI model to generate a custom dataset for the fine-tuned AI 🤯
Been itching to work on this for a bit - and after @Vercel announced new AI integrations last week, I thought it was a good excuse to work on this and play around with the Vercel AI SDK on something new.
I also wanted to spend more time with RSC and server actions, and honestly, it's been awesome. Blown away by what I was able to whip up in just a handful of hours.
So far, this works with @OpenAI, and I'm adding @Replicate + other models/providers that support fine-tuning. Excited to see this come to life!
Check out the (very early) live demo: https://t.co/QKlCyvQbhz