The ways people cope today, and what Unl adds.
Custom instructions, rules files and memory each do part of the job. Unl keeps what you decided and why, and hands it to every AI you use.
| The usual way | Where it stops | With Unl |
|---|---|---|
| Custom instructions in ChatGPT or Claude | Stays inside one app, and applies to every chat whether it's relevant or not | The same decisions reach every AI you use, and only when they bear on the turn |
| CLAUDE.md, .cursorrules, project knowledge | Lives in one repo and one editor, kept up by hand, and goes stale | Your decisions follow you across repos, tools and models, and update the moment you change your mind |
| Memory features | Remember what you said, not why you decided it | Keeps each decision with its reason, and what it replaced |
| Building your own MCP memory server | You host it, configure it and keep it running | Connect once. Unl is the why agent already running |
You stop writing things down for your AI. Unl notices decisions as you make them, and hands each AI the one that matters, when it matters.
Custom instructions in ChatGPT or Claude
Custom instructions travel nowhere
Custom instructions are a good way to tell one app how you like to work. They load into every chat in that app, whether the chat is about your launch or a recipe, and they stay behind when you open another AI.
Unl holds your decisions outside any one app. Connect it to ChatGPT, Claude, Codex, Cursor or any tool that speaks MCP, and each one gets the same decisions. It only hands over the ones that bear on the turn in front of it, so a chat about something else stays clean.
Keep your custom instructions for tone and format. Let Unl carry what you decided.
CLAUDE.md, .cursorrules, project knowledge
A rules file carries the how. Unl carries what stands.
CLAUDE.md, AGENTS.md and .cursorrules earn their place. They are the right home for your stack, your style and your commands, the things that change when the repo changes. They are read once when a session starts, and after that they only know what someone last wrote into them.
What you decided last week, what you tried and ruled out, and where the work stands now change between sessions. Nobody edits the file in time, and a stale line reads exactly like a current one. Unl keeps those decisions with their reasons, marks the old one replaced when you change your mind, and hands the one that bears into the agent before it acts, in every repo, tool and model you use.
Documentation works the same way. A README, a docs folder or a wiki is the right home for reference that holds still, and it is read when you point at it. A decision that has just changed has to arrive on its own.
Keep the file for conventions and for the one line that points your agent at Unl. Let Unl hold the decisions.
Memory features
Memory remembers what came up. Unl keeps where it landed.
ChatGPT's and Claude's memory features are real and useful. They build a picture of you, your preferences, your projects, what you tend to ask for. That picture stays inside the product that built it.
What memory does not hold is the decision. It can recall that pricing came up four times. It cannot tell you which of those ended in the call you made, whether it still stands, or why. Unl keeps the decision, its reason, and what it replaced, and it goes with you to every AI you use.
Keep your AI's memory for who you are. Unl is the layer for why you decided.
Building your own MCP memory server
You could build it. Unl is already running.
Unl reaches your tools over MCP, the open standard your AI tools already speak. So the do-it-yourself version is a memory server of your own. You host it, configure it for each tool, keep it running and decide how it tells a current decision from an old one.
With Unl you connect once and sign in. The why agent is already running, it already knows which decision replaced which, and it already works across the tools you use.
If you already run your own MCP servers, Unl sits beside them as one more.
Plain answers.
01Does Unl replace my CLAUDE.md or rules file?
No. Keep the file for your stack, style and commands. Unl holds the decisions and their reasons, which a file only knows as of its last edit, and hands the one that bears to your agent before it acts.
02ChatGPT and Claude already remember me. What does Unl add?
Memory keeps facts about you inside one product. Unl keeps the decisions you made, why you made them and what they replaced, and brings them to every AI you use.
03Should I turn my AI's memory off if I use Unl?
There is no need. Memory keeps who you are. Unl keeps why you decided. They do different jobs and work side by side.
04What happens when I change my mind?
Say so in the conversation. Unl marks the old decision replaced, keeps it readable with its reason, and from then on hands your AI the new one.
05Could I build this myself with an MCP server?
You could. You would host it, configure it for each tool and keep it running. Unl is the why agent already running, one connection and one sign-in.