What Kit is

Kit is a memory that stays with you. Every conversation you have with it, on your phone, in your editor, in this window, lands in one place, and every new conversation starts from there. You never re-explain. Overnight it sleeps: it reads the day back, connects what belongs together, and quietly checks that what it remembers still matches where it came from.

What Kit can do

• Remember across everything. Tell it something on Telegram, ask about it in Claude Code the next morning. Same Kit.
• Wake with context. A new session opens already knowing what you were in the middle of, what you decided last week, and what it noticed while you were away.
• Answer from memory, and say where from. Ask what you decided, when something happened, who said what. It answers from what it holds and cites the memory. If it does not have it, it says so instead of guessing.
• Notice. Patterns across weeks, things you said you would do, threads that went quiet. It brings them up when they matter.
• Read what you connect. The history of the coding tools you use (Claude Code, Codex, Cursor, OpenCode, Kilo Code and more), a ChatGPT export, your Telegram conversations with it. Each one becomes memory it can draw on.
• Sleep, and keep itself honest. The nightly dream consolidates the day, and Health shows how it went, including how well it can still find what it stored.

How to ask

Kit works best when you talk to it the way you would to someone who was there.

• Name the thing. "The Northgate export we discussed" beats "that thing from before". A name is a handle it can pull on.
• Ask for what you already told it. "What did I say about the invoice terms?" "When did I last talk to Sarah?" This is the point of it. Use it.
• Ask what it noticed. "What have I been circling this week?" "Anything I said I would do and haven't?" It keeps this kind of thing; it will not volunteer all of it unless asked.
• Tell it to hold something. "Remember that Sam is they/them." "Keep this: the launch moved to October." Said plainly, it is kept plainly.
• Correct it. "That's wrong, it was Thursday." A correction is a memory too, and it sticks.
• Ask what it doesn't know. "Do you have anything on the Meyer contract?" A clean "no" is useful; it tells you what to give it.
• Ask about time. "What happened yesterday?" "What did we do in July?" Its memories are dated, and it can walk them.

Two things that help more than they look:

• Short is fine. You do not need to write a prompt. "Kit, the Meyer thing?" works.
• Give it a moment. On Telegram, a longer answer can take a few seconds. If it says it is on it, it is.

Where things live

• Memories is everything it holds, searchable, with the source each one came from.
• Activity is your days: conversations, what it did overnight, what it noticed.
• Data sources is what it is reading, and how each import is going.
• Where you use Kit is the tools it is connected to, and where to connect another.
• Health is how it is doing right now. Green means fine; a row that wants you is amber and opens itself. Under Dreams it says how the last sleep went and how well it can still find what it stored.
• Settings is its name, its model, its sleep, and what it can reach.
• How to is this page.

When something is off

Open Health first. If a row is amber it says what and usually offers a button. If Kit is quiet on Telegram, that is almost always the poller, and Health will say so. If neither explains it, Send us feedback goes straight to us with the details attached, and a screenshot of what you saw is the single most useful thing you can add.

Send us feedback