Clawdis🦞

Claw is the law

Let’s win this thing.

Beta tomorrow.

Just passing over 180 @github stars! We will be reaching out shortly to see if anybody would like to get an early beta test access for Clawdis before we fully roll out. Locked in and pushing for the @BagsApp Hackathon win. github.com/ClawdisAI/Claw…

⏳ We will win the @BagsApp hackathon. Stay Connected. bags.fm/2yDL2okKMtkxtS… github.com/ClawdisAI/Claw…

Now here's what's coming next. → Persistent semantic memory — your agent remembers everything across sessions. "What did we discuss last Tuesday?" just works. → Agent-to-agent delegation — your main agent spins up sub-agents. One researches, one codes, one schedules. All coordinated through the gateway. → Screen awareness — the agent sees what's on your phone and acts on it. "Order me the same thing as last time" while looking at Uber Eats.

I also built the ecosystem around it: → Bird — a CLI so your agents can post to Twitter → QMD — search engine for your docs and knowledge bases → Spogo — Spotify in your terminal → GoGCLI — Gmail, GCal, GDrive from the command line → MCP servers for @X and Obsidian Every tool connects back to one local brain.

Under the hood, for the engineers: → TypeBox + AJV frame validation at the protocol layer → Idempotency cache on every mutation — nothing double-sends, ever → Per-session model selection with custom providers → Sandboxed workspace with personality injection (SOUL.md, AGENTS.md) → Cron scheduler with isolated execution → Group chat activation modes — mention, always, dm-only 2,435 commits. Nothing is a wrapper. Everything is from scratch.

I left a career across McKinsey, Bridgewater, Santander and Motive to build something I actually wanted to use every day. For the past few months I've been heads down building Clawdis — a personal AI assistant that runs entirely on your own hardware & how I think it is perfect for the @bagsapp hackathon. No cloud. No subscription. No compromise. Here's everything it does — and what's coming next 🧵

Skills turn Clawdis into a personal operating system. Twitter bookmarks auto-categorized every 20 minutes. Smart home controlled from a text message. GitHub managed from Telegram. Spotify from terminal. Obsidian searched from anywhere. Tuya. Eufy. Tailscale. Dashlane. Put.io. All written in Go. All pluggable. All local.

Canvas changed everything for me. Most AI replies with text. Clawdis replies with interfaces. The agent decides what you need to see — a dashboard, a form, a visualization — builds it in real-time, and pushes it directly to your phone. Rendered natively. Interactive. Live. Your AI doesn't describe the answer. It shows you.

Then we built Talk Mode. Not voice commands. Not "Hey Siri" followed by a loading spinner. Continuous, natural conversation. Wake word detection runs on-device. Speech-to-text on-device. TTS through @ElevenLabs . You talk. It responds. No button. No delay. Just a conversation with something that actually knows you.

Device pairing works like AirDrop. Your iPhone discovers your Mac over Bonjour. You confirm a 6-digit OTP. That's it. The phone becomes a node — camera, microphone, screen, canvas — all available to the agent. Fully local TCP bridge. No account. No sign-in. No cloud relay.

Surfaces, not apps. WhatsApp on your commute. Voice while you're cooking. Mac when you sit down. Telegram in a group. Discord with your team. One persistent memory across all of them. No sync. No cloud. Just your devices talking to each other over your own network. You don't switch apps. You switch contexts. The AI follows.

The problem with every AI assistant today: you're renting intelligence from someone else's infrastructure. Your conversations? Stored on their servers. Your context? Gone when the session ends. Your data? Not yours. Clawdis flips that entirely. A local WebSocket gateway runs on your machine. That's the brain. Everything connects to it. Nothing leaves your network.

What if your AI assistant didn't live on someone else's server? What if it ran on your Mac. Paired with your phone. Talked to you on WhatsApp, Telegram, Discord — and never sent a single byte to the cloud? I built that. It's called Clawdis. Let me walk you through it.

I left a career across McKinsey, Bridgewater, Santander and Motive to build something I actually wanted to use every day. For the past few months I've been heads down building Clawdis — a personal AI assistant that runs entirely on your own hardware. No cloud. No subscription. No compromise. Here's everything it does — and what's coming next 🧵

@finnbags 👀

Tonight.

@BagsApp 🔥

@ClawdisAI 👏

Officially entered the @bagsapp hackathon with Clawdis. The token was created to sustain ongoing development. Revenue from transactions goes directly into server costs, engineering, and platform expansion. No grants, no investors -- just a product funding itself.