/community/guides/connecting your ai
innernet
docsskillscommunitycompanycomparesecurityprivacytermscontact

connecting your ai · step 3 of 5

what to say to your ai

once a tool is connected there is nothing to learn. the things people say most, what each one does, and what you'll see.

/community/guides/what-to-say-to-your-ai
3 of 5
step 1 of 54 minconnect your ai toolsclaude, chatgpt, cursor, claude code and the rest, each connected once, and how to check it worked.
step 2 of 55 minstart in one commandconnect every ai tool on your machine to one memory, see what the command does at each step, and know what to try the moment it finishes.
step 3 of 54 minwhat to say to your aionce a tool is connected there is nothing to learn. the things people say most, what each one does, and what you'll see.
step 4 of 51 minbring what you already haveyour ai tools have been remembering things in their own files. bring that in once, and let netti sort it.
step 5 of 52 minthe chat windowthe small innernet line inside your ai chats, what it shows when you open it, and how to set it the way you like.

once a tool is connected, there is nothing new to learn. you don't open innernet, call anything by name, or paste anything in. you talk to your ai the way you already do, and it decides when to read your memory and when to add to it.

this page is the phrasebook: the things people say most, what each one does, and what you'll see. every example works in any connected tool, whether that's claude, chatgpt, cursor, claude code, codex or the rest, because they all read the same memory.

first, check it's there

open a new chat. most tools only read their connections when a chat begins, so an old one won't know yet. then say:

the simplest test. any answer at all means it's connected.
the simplest test. any answer at all means it's connected.

if it lists your memories, or says you don't have any yet, it's connected. if it answers as though innernet doesn't exist, restart the tool and try a fresh chat; connect your ai tools has the rest.

ask what's known

ask the way you'd ask a colleague who was in the room. a whole question finds more than a keyword.

  • what did we decide about pricing, and why?
  • in lisbon-app, how do deposits work? name the memory when you keep more than one.
  • what did we use before stripe? ask about the past on purpose, and the history comes with it.
the ai hands your question to innernet word for word, and answers from what comes back.
the ai hands your question to innernet word for word, and answers from what comes back.

innernet doesn't send back your whole memory. it sends the few passages that answer the question, so it stays quick however much you've kept. and when the honest answer is nothing, it says so rather than letting the ai guess. that's how you know the rest really came from your memory. more in ask it anything.

keep something

  • remember that the refund window is 14 days. one line, kept. netti files it into the right page a little later, and it's already there for the next question.
  • save what we decided today to innernet. a whole write-up becomes a page of its own.
  • save this conversation to a new innernet project called "trip to lisbon". starts a new memory from what you've been talking about.
kept at once, and filed into the right page a little later.
kept at once, and filed into the right page a little later.

most of the time you won't have to say it at all. a connected ai notices a decision when one is made and keeps it on its own, guided by the project's capture protocol, and in a code repo every commit is kept for you. see capture without trying.

tell it about you

some things are about you, not about a project: how you like to work, the tools you rely on, the people you work with.

  • remember i like short answers, code first.
  • remember i work best in the mornings.
  • what do you know about how i like to work?

these become facts about you, kept across every project and every tool, so the next tool starts from who you are instead of from zero. you decide who sees each one. a private fact never leaves innernet unless you let an app read it, and every time one does it's written down where you can see it. you, and your private facts shows how.

pick up where you left off

  • at the end of a session: leave a handoff. what got done, and what's next.
  • the next morning, in any tool: where did we leave off?
a handoff, read back the next morning in a different tool.
a handoff, read back the next morning in a different tool.

this is the habit that makes switching tools painless. the handoff doesn't belong to the tool that wrote it; it belongs to the project.

try an idea without losing the old one

  • branch lisbon-app to try usage-based pricing.
  • what changed on that branch? merge it if nothing clashes.
a branch is a copy to think in. the original stays as it was until you merge.
a branch is a copy to think in. the original stays as it was until you merge.

nothing on the original changes while you explore, and if the idea doesn't work out, you simply leave the branch. more in branch your memory.

a few more

saywhat happens
add "call the venue" to my tasksit lands on the same list your dashboard shows
what's on my list today?reads the list back, in order
start a living document for the competitor researcha page that re-organises itself as findings land
find a skill for writing release notessearches the community shelf of skills
what's in the lisbon-app memory so far?a quick tour of a memory's pages

what you never need to say

  • a tool's name. the ai picks the right one. if you watch closely you'll see names like innernet_capture go by; you never have to type them.
  • a key or a password. if an ai ever asks for one to "connect innernet", it's confused. give it the link, https://innernet.live/api/mcp, and let the browser sign-in run.
  • "use innernet" at the start of every chat. once a tool is connected, it's there.

from the terminal

if you connected with npx innernet, the same terminal carries a short version of this page, and a way to check on things:

npx innernet guide     # this phrasebook, and every guide, in the terminal
npx innernet status    # what's connected, and whether your commits are landing
the status check: every tool, this folder's memory, and the last commit that reached it.
the status check: every tool, this folder's memory, and the last commit that reached it.

the whole command, step by step, is start in one command.

next: bring what you already have.

every chapter

the fundamentals5 guidesconnecting your ai5 guidesthe dashboard4 guidesfor builders2 guidessharing3 guides

one memory, every ai tool.

innernet is memory your ai tools read live: every project, every decision, every skill, in one place, connected once.

start your innernetevery guide