Local-first context for any AI

One private AI across Notion, GitHub & Obsidian

One private knowledge graph from all your tools, so any AI can reason across everything you know. On your device. On your terms.

Local-first · Source-available · No data stored or trained on
Notion · GitHub · Obsidian
Project Atlas spec
#214 blocked on a decision
#209 just closed
Daily note, 4 todos
+ synced back to your device
Assembled by Celis
brief this morning

Before you open a tab:

3 GitHub issues blocked on your Notion spec
2pm on Project Atlas, context ready
2 Obsidian todos on issues that just closed
What's blocking the Atlas launch?

The auth model isn't approved, so #214, #221, #233 all wait on it. Decision sits in your Notion doc.

The problem

Your knowledge is scattered. Your AI is blind.

Spec in Notion. Issues in GitHub. Notes in Obsidian. Every answer means five tabs and a dozen searches, and no AI sees across any of it.

Celis changes that.

How it works

Connect, unify, ask

Scattered tools to one private graph, in three steps.

Step 1

Connect

Link Notion, GitHub & Obsidian. Celis pulls your data back to your device, not someone else's cloud.

Step 2

Unify

One private graph that resolves the same task, project, and person across every tool.

Step 3

Ask & act

Ask anything across everything. Get a daily brief. Mark things done, with any model, cloud or local.

Daily brief

Your morning brief, before you open a single tab

Celis opens your day with what matters, assembled from everything you know. Then you just ask.

Today's brief
live
3 GitHub issues blocked on your Notion spec.
2pm on Project Atlas, context ready.
2 Obsidian todos on issues that just closed.
Why we're different

Everyone has the AI. Almost no one has the context.

The model isn't the hard part anymore. The context is. Celis ingests your messy data, links it across tools, and assembles the right context for every question.

Bad context in, useless answer out. Celis gives your AI the context it's missing.

The trust block

Your knowledge stays on your device

Everything runs on your own device, by default and by design.

Local-first. Your graph lives on your machine, not our servers.

You choose the AI. Fast cloud or fully local. Never stored, never trained on.

Total transparency. See exactly what's sent, and where. Every time.

Open & auditable. Source-available. Don't trust us, read the code.

On your device
PA
Project Atlas
acme-corp/atlas
live
SpecNotion · auth model
Blocked#214, #221, #233
DecisionPending in spec
Everything runs here
names & identifiers removed
Shared with a bigger AI
••
[project]
[repo]
SpecNotion · auth model
Blocked#[id] ×3
DecisionPending in spec
Names & identifiers removed first

We literally can't see your data. That's the point.

How it compares

One graph your tools can't give you

Notion AI / ChatGPT Obsidian alone Celis
Sees across all your tools
Your data stays local
Choose any model (cloud/local)
Proactive daily brief
Doesn't replace your tools
Our mission

Your knowledge should work for you, not Big Tech

For years, your data made other people rich. Celis flips that. Your knowledge is yours to use, to think with, and one day, on your own terms, to profit from.

It starts here, on your device.

Early access

Be among the first to think with Celis

We're onboarding a small group of founders, builders, and teams who live across Notion, GitHub & Obsidian.

No spam. Early access, behind-the-scenes build logs, and a say in what we build next.

FAQ

Questions, answered

How does it work?

Celis connects once to Notion, GitHub, and Obsidian and pulls your data back to your device. It builds one private graph that resolves the same task, project, and person across tools. Then you ask anything, get a daily brief, and mark things done. Nothing leaves your device unless you choose to send it.

Where is my data stored?

On your own device. Your graph lives with you, not our servers, and we can't see it. When you use a cloud model, Celis masks names and identifiers first and shows you exactly what's sent.

Can I use it with ChatGPT, Claude, and local models?

Yes. Celis is model-agnostic. It assembles the right context for every question and hands it to whatever model you pick: fast cloud or fully local. Per source, per question.

Does Celis replace my tools?

No. Notion, GitHub, and Obsidian stay where they are. Celis sits underneath as the context layer that lets one AI reason across all of them.

Is it really private?

Yes, by design. Local-first by default, source-available so you can read the code, and total transparency on anything sent to a cloud model. Don't trust us, audit it.

What happens if Celis shuts down?

You keep everything. Your graph is stored in an open format that doesn't depend on us, and the source is available, so it keeps working. We're independent and building this to last.