Skip to content

Noetum is in private development. Help shape what comes next.Request early access

noetum

Personal AI memory that cites the original.

Noetum keeps what you save, write and decide as a private record. Claude answers from it, marks what it inferred, and links every answer to the exact passage it came from.

Request early accessSee it answer
Illustration inspired by Noetum’s Brain Canvas: fine cyan and amber connections form a map of related memories.

A question for your memory

“What changed since I made that plan?”

You changed your mind. Your memory should keep the reason, too.

Follow an answer to its source
Brain Canvas, illustrated. Explore a sample answer below.

Running today on its founder’s own archive of more than 15,000 saved posts and everything around them

  • X likes
  • Pinterest
  • WhatsApp
  • Web pages
  • Files
  • Notes
  • Journal

Every answer points to the original

Ask Jarvis a question. Claude answers from your records, numbers the ones it used, and marks what it inferred. Each number opens the note you actually wrote.

You asked Jarvis

Am I still planning to work from Lisbon in October?

No. You first planned to spend October in Lisbon. On 22 September, you changed the plan to stay in Istanbul for the project.

Inferred from the two records below

Select a citation to read the original
Source 1 · Your noteDemo

Original plan

“October: work from Lisbon for a month.”

A note about where to work next month.

Earlier plan · kept in the history

Source 2 · Your noteDemo

Change of plan

“Change of plan: I’ll stay in Istanbul for October. The project needs me here.”

A later note about the same decision.

Later note · original wording preserved

Demo · Explore an answer and its original sources.

How Noetum keeps a memory honest

The original stays intact. Your corrections stay in the history. The map helps you explore how it all connects.

Noetum’s Brain Canvas, with 800 synthetic records connected across three regions and a list-view alternative.
The app’s Brain Canvas. Every view also has a plain list. Product visualization.
Originals are never edited

Records are stored exactly as they arrived. A correction is a new entry with its own date, and the earlier version stays in the history.

Every statement says where it came from

Recorded by you, inferred by Claude, or corrected by you. Anything a model derives carries its source, the model and the extraction version.

Four kinds of time

When something happened, when you saved it, when Noetum recorded it and how long it stayed true. “I lived in Ankara” and “I live in Istanbul” can both be right.

Views are rebuilt from the record

Search, the knowledge graph and the 3D Brain Canvas are projections. They can be rebuilt at any time, and each has a list or timeline alternative.

Nothing acts without asking

Agents read through a bounded gateway. Sending, deleting or anything outside Noetum waits for your approval and is written to an append-only audit log.

Where Claude does the work

Noetum keeps the record, the history and the permissions. Claude reads it, labels it and answers from it.

Classifies and tags
Claude reads each incoming record and proposes topics, people and things. Every label is stored as an inference you can confirm or correct.
Answers in Jarvis
Noetum’s in-app assistant answers with Claude and cites the records it used. A claim without a source is labeled as reasoning.
Reads memory over MCP
MCP, the Model Context Protocol, is how Claude connects to outside tools. Claude Code uses Noetum’s MCP server to search memory, read sources and check quotes.
Builds Noetum
The product itself is built and verified with Claude Code agents, against written plans and tests.
Claude Codeconnected to Noetum over MCP

Why did I put the launch on hold?

  1. list_memory_candidates“launch on hold”2 records
  2. read_memory_sourceFriday project reviewmeeting note, 18 Sep
  3. validate_memory_evidencequoted passageexact match

You paused the launch to fix onboarding first. In your Friday project review you wrote: fix onboarding before we invite more people.

Demo · Claude Code connected to Noetum
Portrait of Emre Canik, founder of Noetum.

Built on one real archive first

Emre Canik started Noetum with a personal archive of saved posts, pins, chats and notes: too many to find again, and too many to trust a summary of. Noetum runs on that archive today, built with Claude Code.

15,000+saved X posts in that archive, with media, translations, topics and people, next to Pinterest, WhatsApp, web pages, files, notes and a journal.
More about Emre

Questions

Can I use Noetum today?
Noetum is in private development and currently runs locally for its founder. Get in touch to discuss early access and the records you would like it to remember.
How is this different from chatting with an AI?
Noetum keeps a durable memory outside the conversation: original sources, dates, relationships and your corrections. The assistant can return to that evidence later, so context does not have to start again with each chat.
Does my information stay on my machine?
The current app stores your records locally. AI features can send selected context to external model providers after a data-policy check. Local storage does not mean every AI operation runs offline. You control the configured providers and permitted data access.
What can I explore in the demo?
Select either citation to compare an earlier plan with a later correction. The demo uses example notes so you can explore the experience before requesting access.

Request early access

“Here’s what I’d like my AI to remember:”

Finish that sentence in an email to Emre. Tell us what you want to remember and help shape Noetum’s early access.

Request early access