innernet has been one memory you carry between every ai you use. 2.0 takes it further: into the agents you build, into chatgpt in one press, into how every ai talks to you, and into a netti you can ask anything. all of it is open today, on every plan.
added
agents that remembergive an agent you build a memory of its own. paste one block into the ai you code with and it adds two calls: what the agent reads before it answers, and what it keeps after. from the next turn it remembers everyone it serves. each agent gets a page of its own with four spaces, memory, access, activity and settings, so you can always see what it knows and what it did.
one access levelone setting decides how far an agent may reach into your own memory. sealed reads nothing of you; asks first, the default, holds what it learns about you until you keep it or let it go; trusted reads your personal memory and files what it learns on its own.
an agent, made by askingask the ai you already talk to, and it makes the agent with its key, wires the memory into your code and shows you what the agent kept. the same works from a terminal with npx innernet agents, and over plain https with the agents api.
the innernet plugin for chatgptinnernet is a published plugin in chatgpt. open it, press connect and sign in: that is all of it. it works on every chatgpt plan, free included, with no developer mode and no link or key to paste. add it once on chatgpt.com and the phone app has it too, and codex carries the same plugin.
innernet in every ai you useone memory in the places you already talk to an ai: chatgpt and codex through the innernet plugin, claude, grok and grok bot, perplexity, le chat, gemini, notion, and the agents in your editor and terminal, cursor and vs code among them. each one is a press or a paste away, every one signs in through your browser, and each has its own guide at innernet.live/docs/everywhere.
ask nettitalk to netti about anything you keep, right on your dashboard. it answers from your own memory, and it can act on it too: rename a page, move what landed in the wrong place, rewrite a section, park a direction, or let a fact go.
connect with nettipress connect on any tool and netti walks you through it in a conversation: what to paste and where, a live check that hears the tool's first call, and what to connect next.
moving inbring in everything you already have, two ways. in your ai: one prompt, and it moves its own memory of you over. or in a file: a chatgpt or claude export, notes, a folder. netti sorts it as it lands, and one sheet shows you what moved in.
your agent identitytell innernet how you want an ai to talk to you, answer you and behave, or let netti notice it when you tell any ai how you like to work. you keep each line with one press, and every model that reads your memory gets the same one: claude, chatgpt and cursor all answer the way you asked.
netti-1, the person modelthe memory as our lab measured it, on for everyone. when an ai asks, it is handed the facts that bear on the question, best first. each night netti folds the day you had into a reading of how you work, kept sealed under your own key, and an app only ever receives what you have cleared it for. innernet's agent memory answers 89.4% of longmemeval-s on the same code your agents call, graded with the test's own judge prompt on claude sonnet 5.5 rather than gpt-4o.
the daily digestyour day, sent back to you by email: what you covered, what you decided, and where each new thing netti learned about you went. you choose when it lands, and a minute's look keeps netti filing things the way you would. the mail is your own switch.
innernet for developersone page for building with innernet: the three ways in, memory for the agents you build, the api, the research behind it, and what it costs. at innernet.live/developers.