This article was originally published on the Vilix AI blog.
I run my whole operation from my phone. Not as a figure of speech. Every AI tool I use shares one memory through Vilix AI: Claude Code on my laptop, Codex, the Hermes agent running 24/7 on an Azure virtual machine, Claude and ChatGPT on my phone. I connect each tool once, and they all know everything.
Connect once, every tool knows everything
Each AI client connects to the same Vilix AI account over MCP. I plan in one tool, build in another, and the context follows me. No copy-pasting prompts between apps. No re-explaining the project to every new assistant. Project rules, tasks, decisions, and preferences live in one shared layer, so every tool reads the same truth. When two tools save conflicting info, the last write wins, which means I correct something once and it is corrected everywhere.
It stores conversations, not just facts
Vilix AI saves the full exchanges, with timestamps and source attribution. That matters more than it sounds. I can ask any connected tool when we discussed something, or what we decided months ago, and it knows. Things I have completely forgotten are still there, with the date attached. It is the difference between a notes file and a memory.
The cloud part is the real superpower
The memory lives in the cloud, not on my laptop. I manage zero infrastructure. Because of that, my phone becomes a remote control for all of it. I get an idea anywhere, I tell an AI on my phone, and the work happens. Meanwhile Hermes keeps building on the VM around the clock with Codex and Claude Code, all three reading and writing to the same memory I just updated from my phone.
A real example: my website
This blog is the demo. From my phone I can say "add this FAQ to my website" or "add this comparison page" or "add the Product Hunt badge", and it gets done. The agents know my site, my accounts, and my tools, because Gmail, Notion, and everything else are connected through the same MCP setup. I connected each account once. Every AI I use can reach them.
All of those tool connections live in Composio. It holds my MCP tools, Gmail, Notion, and everything else, in one place. So when I connect an account once in Composio, every AI tool I use can reach it. I never wire up the same integration twice.
It runs my content machine too
Content creation works the same way. Instagram, X, and LinkedIn are all connected, so I can publish from my phone with every tool understanding the full picture. If I get a post idea, I just say it. If I want a slideshow for Instagram, it gets made. I can record a video on my phone and have Claude Code use Remotion on the Hermes agent to edit it. Canva handles design. And scheduled jobs run in the background doing research and finding content ideas automatically. The whole marketing loop runs on the same shared memory.
Skills mean no one ever re-learns anything
Vilix AI also hosts skills: reusable procedures that any connected tool can follow. Once one AI learns how to do something step by step, that procedure is saved, and every tool knows it from then on. Nobody re-learns, nobody forgets. The workflow itself becomes part of the memory.
Projects and tasks keep every build organized
Everything I work on lives in projects, and every project has tasks with the details attached. I plan on my phone, from the car, from a coffee shop, wherever. We talk it through, I agree on the plan, and then it is simple: create a task and hand it to Claude Code or Codex to build it. The builder picks up the full context, the plan, the decisions, everything.
I switch tools by strength, and context comes along
Different models are good at different jobs, so I switch freely. Claude is excellent at frontend and design. Codex handles backend work. On my phone I often reach for Gemini or ChatGPT because they answer fast, while Claude takes more time there. If I hit a rate limit, I just move to another tool. The context always comes with me, so switching costs nothing.
Under the hood: retrieval, not dumping
When an AI needs context, Vilix AI retrieves the memories relevant to that moment instead of dumping the whole archive. It runs on RAG over a vector database, so recall is semantic: it finds what you meant, not just what you typed. Your memory keeps growing, and the right piece always surfaces at the right time.
It is yours: export and delete
You can export all of your memory in a portable format anytime. If you ever move to another provider, your context comes with you. You can delete a single memory or wipe the entire account instantly, no waiting period. Your data, your call.
FAQ
Do I have to set up each AI tool separately?
Yes, each client needs its own one-time connection to your Vilix AI account. After that, they all share the same memory.
Does connecting a tool import its old chat history?
No. Vilix AI starts recording from the moment you connect. Old chats stay where they are.
What happens when two tools disagree?
Last write wins. The most recently saved memory overrides the older one, and retrieval is recency-aware, so every tool sees the newest version.
Which tools work with it?
Claude, Codex, Cursor, OpenClaw, Hermes, Grok, Manus AI, GitHub Copilot, Windsurf, Lovable, Muse, and any MCP-compatible AI.
Can I really run it all from my phone?
That is the whole point. The memory is hosted in the cloud, so any connected AI on your phone reads and writes the same context as the agents running on your laptop or a cloud VM.
If you want this setup for yourself: try Vilix Pro free for 7 days, no credit card.
Top comments (0)