START WITH ONE WORKFLOW — GIVE AI THE CONTEXT IT IS MISSING
    Penumbra

    Ontology Compiler · v0.1

    The capability
    library.

    Penumbra is built from a small set of typed primitives. This page is the source of truth for what they are, what they do, and what we mean when we use them — so you and your agents can speak the same language.

    The map

    Four layers,
    thirteen primitives.

    The system is small on purpose. Define the Shape. Turn on the surfaces. Operate through them with provenance. Reconcile what compounds back into the ontology.

    01 · LayerEncoding LayerHow expert judgment becomes typed, governed primitives instead of tribal knowledge.3 primitives ↓02 · LayerProject LayerHow each new piece of work inherits your domain context on day one.3 primitives ↓03 · LayerDelivery LayerHow agents and humans produce work that sounds like you — with receipts.4 primitives ↓04 · LayerCompounding LayerHow every project makes the next one smarter.3 primitives ↓

    Layer 01

    Encoding Layer

    How expert judgment becomes typed, governed primitives instead of tribal knowledge.

    01

    Shape

    n.
    PRIMITIVE

    A typed unit of expert judgment.

    What it is
    A Shape is a structured definition of one thing your organization knows how to think about — a framework, a rubric, a decision rule, a deliverable archetype. It has fields, constraints, and provenance.
    Why it matters
    Without Shapes, your method exists only in heads and slide decks. Shapes make method addressable, composable, and auditable.
    02

    Ontology

    n.
    DOMAIN

    The graph of Shapes that forms your domain model.

    What it is
    An Ontology is the live, versioned domain map of how your organization sees its work — entities, relationships, vocabularies, and the rules that bind them. Each project inherits from it; each project enriches it.
    Why it matters
    An organization without an ontology re-explains itself every Monday. An organization with one compounds.
    03

    Encoding Session

    n.
    PROCESS

    A structured interview that turns tacit method into Shapes.

    What it is
    A facilitated session — sometimes human-led, sometimes agent-assisted — that surfaces what an expert actually does and writes it down in a form the system can run.
    Why it matters
    Most organizations have never seen their own method written down. Encoding is the first time it becomes legible to anyone but its author.

    Layer 02

    Project Layer

    How each new piece of work inherits your domain context on day one.

    04

    Project Context

    n.
    INSTANCE

    An ontology, populated for one context.

    What it is
    When a project begins, your Ontology is instantiated against the specific reality — the entities, the language, the constraints at hand. The result is working context for that project.
    Why it matters
    Discovery becomes a structured walk through your ontology rather than a blank doc and a prayer.
    05

    Project Memory

    n.
    STORE

    Everything the project learned, kept queryable.

    What it is
    A typed, versioned record of every artifact, decision, and revision produced during a project — bound to the Project Context that produced it.
    Why it matters
    Project two starts smarter than project one. Project ten is a moat.
    06

    Discovery

    n.
    PHASE

    Structured ontology-walk, not a blank doc.

    What it is
    The first phase of a project: a guided traversal of your Ontology against the situation at hand, producing the initial Project Context.
    Why it matters
    Discovery is where most projects leak the most judgment for the least return. Penumbra makes it the most leveraged hour of the project.

    Layer 03

    Delivery Layer

    How agents and humans produce work that sounds like you — with receipts.

    07

    Knowledge Agent

    n.
    ACTOR

    An agent bound to your ontology and method.

    What it is
    An LLM-backed actor that drafts, reviews, or operates only through your Shapes and Project Contexts. It cannot freelance outside the ontology.
    Why it matters
    Generic agents give you the median answer. Knowledge agents give you your answer.
    08

    MCP Surface

    n.
    INTERFACE

    The protocol-level surface agents use to read and write the ontology.

    What it is
    A Model Context Protocol interface that exposes Shapes, Project Contexts, and Project Memory to any compliant client — Claude, Cursor, internal tools — without flattening them into prose.
    Why it matters
    Your method should be a first-class API surface, not a copy-paste prompt.
    09

    Provenance

    n.
    PROPERTY

    Every output traces back to the Shape that shaped it.

    What it is
    A structural property of all delivery artifacts: each claim, draft, or recommendation is linked to the framework, decision rule, or precedent it derives from.
    Why it matters
    Reviewability without manual audit. Trust without theatre.
    10

    Review Standard

    n.
    RULE

    What “good” means in your organization, encoded.

    What it is
    A typed criterion that gates whether an artifact ships. Applied uniformly to human, team member, and agent output.
    Why it matters
    Quality stops being one expert's mood and becomes infrastructure.

    Layer 04

    Compounding Layer

    How every project makes the next one smarter.

    11

    Lineage

    n.
    GRAPH

    The directed history of every artifact, decision, and Shape.

    What it is
    A queryable graph linking each output to its inputs, its author (human or agent), and the version of the ontology that produced it.
    Why it matters
    Every project becomes a corpus you can interrogate, not a folder you forget.
    12

    Compounding

    n.
    EFFECT

    The thesis. The whole point.

    What it is
    The structural property of an organization whose ontology, memory, and review standards improve with each project instead of resetting.
    Why it matters
    Most organizations are linear. Penumbra organizations compound.
    13

    Governance

    n.
    POLICY

    Who can change the ontology, and how.

    What it is
    Versioning, review, and access policy applied to Shapes and the Ontology itself. Experts promote; agents propose; nothing ships unreviewed.
    Why it matters
    An ontology without governance becomes a wiki. A wiki is where method goes to die.

    Glossary

    A–Z index

    Use the domain

    Want to see one Shape turn into a working surface?

    The research preview is hands-on. We start with one slice of your domain, map the first Shapes, and show what they can power.

    BOOK A WORKING SESSION →Or see builder use cases →