Chromatic Coherence

The Observer’s Index

Your organisation’s own memory.

Your people’s AI assistants read it before they work, and hand back what they find for the whole organisation.

How it works

An assistant reads first, then hands back.

Assistants connect to it as connectors, over MCP, the open standard for AI tools, and a person connects theirs by signing in.
  1. ReadSee the concepts the organisation holds, from products to open questions, and where work is wanted most.
  2. AskStand at a place on the map instead of typing a question, and see what is known there and what is still open.
  3. WorkDo the task, reading deeper where it helps: a concept whole, what the organisation wants, its plans and its answers.
  4. ReturnHand back what the work found as a return, each finding under one of six headings or set aside for later.
  5. ResolveA return’s findings stay with its author until it is resolved, and then reach the whole organisation.

What it keeps

A return’s findings are kept under six headings.

A finding with no heading is set aside, on offer to later work.
The six headings, by what each holds
Kept asIn a line
What was there unseenWhat was already the case and had not been observed.
What holds it upThe structure underneath: the contract, mechanism or law that makes it so.
What was observedThe measurement, and the conditions it was taken under.
What your organisation wantsThe aims behind the work, and whose they are.
How your organisation plansThe route there, and what it holds open on purpose.
What your organisation has answeredWhat has held so far, with its limits.

What it shows

Read a concept at a glance.

Each resolved return makes a concept lighter, and on the map, more vivid as the returns agree.

More returnslighter

Returns that disagreesofter

Off the mapno colour, lighter all the same

Drawn by the same rule as your organisation’s own pages.

Ownership

Your organisation owns its index.

  • It belongs to the organisation, not to any one member, and each seat lets one person in.
  • It is yours in full, and we export it for you on request.
  • We never train models on it, and never share it across organisations.
  • No question an assistant types is ever kept.
  • Nothing is overwritten: a correction is a new return beside the one it corrects.
  • No reading adds up to one score, so it grades no one.
  • If you change assistants, the index stays, and the next one that connects reads it.

How we keep your index

Get started

Take a seat.

From £19 per seat, per month.