Request a demo (opens in a new tab)
Back to the stack

01 · The context layer

Canon

The company's shared memory. Every PRD, ADR, decision, and research note in one corpus, indexed and ready to answer in plain language.

Request a demo (opens in a new tab) See it work
Layer
Context, the base of the stack
Reads
PRD · ADR · DEC · NFR · STD · RES
Feeds
Habitat, Olympus, every agent
Status
Ships with the engagement

The problem it removes

Knowledge scatters. Every stage stops to re-ask for it.

The decisions are in a wiki, a thread, a doc, and three people's heads. So planning interrupts engineering, engineering interrupts the architect, and the answer drifts a little each time it is retold. Nothing is wrong with any one tool. The memory is just spread too thin to trust, and an AI-first team feels it most: agents work from whatever context they happened to be handed.

See it work

Ask a question. Get a cited answer.

A live example

The first answer cites the decision it stands on. The follow-up is instant, because the canon already knows the path.

Why it's different

The base everything else reads from.

Canon sits at the bottom. Habitat builds the workflow on top of it, and Olympus' governed agents act inside that workflow, all working from the one corpus underneath.

Tools that keep their own memory splinter: each holds a partial, stale picture and asks for context again. A shared substrate compounds instead. Every answer it captures makes the next one cheaper, for a person or an agent alike.

See it against your own canon (opens in a new tab)

What it does

One memory, open to the whole team.

  1. 01

    Capture

    Imports the corpus

    PRDs, ADRs, decisions, research, and standards, pulled straight from the repo. One shared memory, not another tool to keep in sync by hand.

  2. 02

    Ask

    Answers in plain language

    Ask a question the way you would ask a teammate. The reply comes back with the documents it stands on, the PRD or ADR behind it, never a guess.

  3. 03

    Keep

    Holds every decision

    Decisions are captured once and kept immutable. When one is replaced, the supersession is explicit. The record never quietly rewrites itself.

  4. 04

    Trace

    Shows the whole graph

    Requirement to decision to architecture to owner. Every link resolves, so an answer can be followed to what justifies it, not just taken on faith.

  5. 05

    Compound

    Feeds the whole stack

    Habitat and Olympus draw on the same canon the team does. Each starts from what is already settled. Nobody re-collects context the corpus already holds.

A closer look

Every answer shows its work.

A sourced answer. The claim and the documents it rests on arrive together, so an evaluator can verify it, not just trust it.
The corpus graph. A decision walks back to the requirement and architecture that drove it. Nothing is a mystery box.
PRD

Product requirements

By module, from capability down to task.

ADR

Architecture decisions

Immutable once accepted, with consequences.

DEC

Lightweight decisions

Day-to-day calls and approvals, owned and dated.

NFR

Non-functional targets

Performance, availability, cost, with SLIs.

STD

Standards and guidelines

The binding rules and the recommended ones.

RES

Research and threads

Discovery notes and the review conversations.

By design

What the corpus guarantees.

Enforced on every run, not promised: a citation is attached to every answer by construction, and one corpus is what the whole stack reads, so no stage works from a private copy that drifts.

See it against your own canon.

Thirty minutes with the people who build it: a working walkthrough and straight answers.

It ships with the engagement, tuned to your stack. The source is yours to keep. No per-seat metering.

Request a demo (opens in a new tab) Start a conversation (opens in a new tab)