---
title: Concepts
order: 2
---
# Concepts

## The mount

Work happens through the mount. An agent (over MCP) or a runner (over plain
HTTP) connects to one organization and gets everything through that single
door: `boot` for the entry documents, `find_process` for routing,
`read_node` and `search` for knowledge, runs for visibility, asks for
decisions. Who you are on the mount (your role, your organization) comes
from your invite code, and a tool your role may not use is a tool your
session never sees. Agents may hold a disposable workspace while working;
nothing durable lives there.

## The company file

The organization is a folder of plain markdown in the owner's own repository:
`ORG.md` (what it is), `AUTHORITY.md` (what may happen, access is not
permission), `processes/` (how kinds of work are done), `records/` (the nouns
the business touches), `decisions/` (every ruling, verbatim), `lessons/`
(what it learned). Delete every harness and app binding and the organization
still runs. That is the guarantee the rest of this page serves.

## Projection

What you read on a mount is a **projection** of the company file: derived,
disposable, rebuildable from the repository at any commit. Every answer
carries the commit it reflects, so any claim can be traced to real bytes in
git. If something looks missing or stale, that is a fact worth reporting. The repository is the truth; the projection is only a view of it.

## Search

One search runs two layers (lexical (BM25) and semantic (embeddings)) fuses
them, and reranks the shortlist with a model that reads the query and each
finalist together. If the semantic layer is unavailable, search degrades to
lexical rather than failing.

## Runs

A run is the unit of visible work: who is working, on what, right now. Real
work opens a run, heartbeats as it moves, and closes with an honest terminal
status, `landed`, `awaiting-ruling` (parked on a human decision; not over),
`conflict` (the target moved; rebuild), or `failed`. A run that stops
heartbeating shows as *stalled*, a fact derived from the clock, not a verdict.

## Decisions

An agent asks; it does not decide. A decision that belongs to the Founder
becomes one link: one question in plain words, a short list of what becomes
true on yes, the evidence one click away. Holding the link is holding the
right to rule. It is made to be pasted into a chat and answered from a
phone. The ruling is recorded in the ruler's own words.

## Roles

An invite code maps a person (or a cron actor) to a role on one organization:

- **observer**: reads everything, changes nothing.
- **steward**: reads, runs work, reports events, asks the Founder.
- **founder**: all of that, plus the rulings are theirs.

Roles are enforced at the tool layer: a tool your role may not use is a tool
your session never sees.

## Tenants

Every organization is a tenant with its own projection, runs, events,
decisions, and members. Reads and writes are scoped to one tenant; a mount
sees exactly one organization. The vendor's own org runs as a tenant in the
same system: the isolation that protects you is the isolation the vendor
depends on.