VOOL now updates itself. One press, and you're on the new version with every chat still there.
Using VOOL on your home network? Other devices can chat, but your files, keys and payments never leave your computer.
Plus 53 ready skills to browse.
VOOL v0.7.5-beta is out.
• Opt-in in-app updates on Mac and Linux; Linux rolls back a failed update by itself
• LAN mode: other devices get plain chat. Files, keys, tools, payments stay local
• Research won't follow links into localhost or the LAN
github.com/Parad0x-Labs/vool…
Using OpenClaw or Hermes? Give VOOL a try.
We're not a fork of either. VOOL is our own agent runtime, built from our original NULLA work.
Our public head-to-head benchmarks show strong wins in memory, quality and token efficiency.
github.com/Parad0x-Labs/vool
VOOL SKILLS -.- Bare-Bones Edition
It's functional, and that's what matters most right now. We'll keep improving it throughout the week.
Go ahead and give it a try. Use one of the available skills, or create and upload your own. Easy as that.
vool.dev/codex/skills/
Give VOOL a goal. Go make coffee. Come back to the steps done. ☕️
0.7.3 is here: /goal, a forget that really clears its notes, and an Activity panel that talks like a human.
VOOL v0.7.3 landed.
Goals that VOOL works through step by step. More commands from chat, shown before they run. Forget that erases from the memory file. Activity in plain words. Smarter Auto model pick.
Your agent. Your keys. Your rules.
github.com/Parad0x-Labs/vool…
VOOL v0.7.2 is out.
X from any chat. USDT sending. Watch alerts and trader tools. Bring your chats from other AI apps. Private chat in 17 languages. Try again, edit, Light and Dark.
Your agent. Your keys. Your rules.
Mac and Linux: github.com/Parad0x-Labs/vool…
Vool retweeted
VOOL v0.7.2 is in the pipeline, and it's a big one.
Coming: X from any chat, USDT sending, watch alerts, trader tools, chat import from other AI apps, private chat in 17 languages, Light and Dark mode, Try again and Edit, and much more.
Your agent. Your keys. Your rules.
Good video.
As VOOL's entire engineering department, we'd add one more step:
step 6 → Make sure the agent can't do what it was never allowed to do in the first place.
Models → Agents → Permissions → Receipts → Memory → Permissions
That matters.
github.com/Parad0x-Labs/vool
Anthropic hired this engineer at $250K–$750K a year because he knows how to build harnesses for multi-agent systems
In a 15-minute workshop, he shows exactly how to build one from scratch
AI → Agents → Harness → Loops → Graphs
step 1 → start with the Claude Agent SDK — the harness handles loops, context, and sandboxing
step 2 → separate the brain from the hands — reasoning in one place, tools in a sandbox, 60% faster to first token
step 3 → run it server-side and log every step — close your laptop and it keeps running, crashes resume from the log
step 4 → make failure cheap — retry dead sandboxes and replay lost context instead of starting over
step 5 → turn yesterday's logs into new memory and skills — the harness wakes up smarter
Anthropic calls this "dreaming"
Most people spend weeks building this by hand
You don't have to
Bookmark and watch it
Then read the full Opus 5.5 harness engineering guide below ↓
And so it begins 🤓
VOOL beta is out. Your own AI, on your own computer. Free and open source, for Mac and Linux.
And this is only the beginning. v0.7.2 is coming very soon: make your own skills right in a chat, and share them with everyone in the VOOL Skill Codex.
Behind the nice extras, the real work never stops: the core runtime, tools, orchestration and memory. All of it is tuned to what we call the Five Weights:
Quality: the answer you get is what matters most.
Speed: nobody wants to wait on a simple task.
Consumption: no wasted tokens. More tokens don't make a better answer, and our early tests already show it.
Context: it follows you, so you never repeat yourself.
Memory: the right facts for this chat, not just more of them.
Every update and addition gets posted here on @vool_dev as it happens. Help us build VOOL better and faster: drop your feedback and skill requests below.
Download it:
vool.dev
github.com/Parad0x-Labs/vool
Vool retweeted
🎙️ VOOL AMA — Sunday, 5 PM UTC
We’re taking it to X Spaces.
This won’t be a one-way presentation — bring your questions, ideas, criticism, feedback, and anything you want to throw at us.
We’ll talk about what we’ve built, what we’re building next, the problems we’re still fighting, and where VOOL is heading.
No polished presentation script. Actual two-way conversation.
See you Sunday. 🫡
Vool retweeted
We’ll be building VOOL in the Crypto World’s Fair Hackathon, powered by @Colosseum.
VOOL is an everyday AI runtime first:
🧠 Local or cloud models
💾 Persistent context + private local memory
🛠️ Tools, skills, plugins and agents
Crypto is optional.
Wallet integrations are OFF by default. If users want x402 payments, Web0 identity, private settlement or on-chain actions, they enable them.
That is where VOOL gets interesting.
Agents can pay for APIs, tools, data or other agents. Skills and plugins can add entirely new capabilities.
But they do NOT inherit authority over your wallet.
🔐 Transaction meaning
🔐 Permission scope
🔐 Spending limits
🔐 Signing policy
These stay inside VOOL’s native runtime — outside the model, skill and plugin layer.
A compromised plugin, rogue agent or malicious prompt injection should not be able to simply rewrite the rules and empty a wallet.
Skills can propose.
Agents can act.
VOOL remains the enforcement layer.
Our Web0 layer takes this further:
john.null → thomas.null
Encrypted on-chain communication, identity, attached payments and agent interactions — without forcing the user into a separate “crypto app” experience.
That is the philosophy behind VOOL:
Build something people actually want to use first.
Then let crypto, x402 and on-chain ownership become invisible infrastructure when they genuinely make the product better.
🧠 AI first
⛓️ Crypto when useful
🔐 User rules always
colosseum.com/hackathon/soci…
