Context is everything.
The graph that remembers who, what, and why — so you don't have to explain it again.
Models are capable.
The context handoff is broken.
Watch anyone work with AI on a non-trivial task. Request. Generic output. Correction. More correction. Each round is you discovering what context was missing and re-explaining it — long, repetitive, and it happens every single time. The tool doesn't fail because it lacks capability. It fails because it never had context to begin with.
Your own thinking isn't flat. There's a sense of self, a handful of contexts you actually work within, and a dense mesh of people, tools and ideas underneath them. Most tools hand a model a flat pile of notes and hope. We give it the shape instead — structured, not stuffed. A briefing, not a bag of chunks.
Built from how you actually work, not a demo.
Wiki pages that link
Every person, project, and idea gets a real page — headed, cross-linked, easy to wander through.
Relationships, not just connections
Edges carry their own story: how you work together, what you're building, how you actually get on.
Contexts that emerge
When a cluster of people and projects genuinely forms its own world, the graph notices — and gives it a home.
Pulls from where you already are
Slack, Gmail, Drive, calendar — connect a source once and it folds straight into the graph.
Chat that already knows your world
Ask it anything, or let it interview you to fill in what's missing.
Your graph, in the open
Every page is yours to read, edit, and wander — nothing about your own context is a black box.
Your graph, read by any agent you already use.
Subtext speaks MCP. Mint a key from Connections, choose what it can see, and paste the endpoint into Claude, Cursor, or any MCP-compatible tool — it reads straight from your graph. No export, no re-explaining yourself in a new tab.
- 01 Create a key in Connections and scope what it can read.
- 02 Copy the endpoint — shown once, yours to keep.
- 03 Paste it into your tool's MCP settings. Done.
{
"mcpServers": {
"Subtext": {
"url": "https://your-engine.example.com/mcp",
"headers": {
"Authorization": "Bearer cek_9f2a..."
}
}
}
}
The richest information lives in the edges, not the nodes.
A relationship is never one thing.
How you work together, what you're building, how you get on — three different lenses on the same connection, and they don't have to agree.
Clusters become contexts.
When enough of those connections cohere into something real, it earns its own place in the graph — not because you filed it there, but because the graph noticed.
Enough surfaces enclose, and a self forms.
One level up from contexts sits your volume — an honest model of what all of this actually means to you, not just what you filed. It's the one part of the graph allowed to disagree with you. Line enough of these up across a team and you get a hypervolume — shared alignment, not just shared documents.
For anyone whose context lives in too many places.
Founders & operators
Clients, hires, decisions — all in one place, instead of scattered across docs and DMs.
Researchers & writers
Sources that compile themselves into something navigable, not another pile of PDFs.
Anyone tired of repeating themselves
To a chatbot, a new hire, or yourself in six months.
A dollar a day. Cancel anytime.
One subscription, no seats. $1/day, billed monthly. Every day resets a fresh credit allowance for ingestion, chat, and graph upkeep — it doesn't roll over, so what it costs you never gets away from you.
20,000 credits every day — real ingestion and chat, not a teaser.
Once you're subscribed, top up with a token pack any time, or turn on auto-top-up and stop thinking about it.
A top-up for the odd heavier day — connect one more source, catch up on chat.
For weeks the graph is doing real work and $1/day isn't quite enough.
Heavy ingestion, heavy chat, agents pulling from your graph all day.
Give your tools
the context they deserve.
Sign up, connect a source or two, and your graph starts building itself.