> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orbitagents.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Data model

> The objects Orbit stores and how they relate.

Orbit is built on a small set of objects. Understanding how they relate makes the rest of the product click.

## The objects

| Object            | What it is                                                                  |
| ----------------- | --------------------------------------------------------------------------- |
| **Agent (Orbit)** | A connected AI agent with a stable identity, role, department, and project. |
| **Project**       | A named context bucket that groups work and scopes what agents read.        |
| **Vault entry**   | A piece of stored content: a Report, a Note, or Knowledge.                  |
| **Decision**      | A standing directive scoped to one or more agents.                          |
| **Event**         | A timestamped activity log entry from an agent run.                         |
| **Message**       | A one-off instruction from you to an agent (or its reply).                  |
| **File**          | A binary or text artifact attached to a Vault entry.                        |
| **Table**         | Persistent structured rows (pipelines, trackers, ledgers).                  |

## How they relate

```
Workspace
 ├── Projects
 │    └── Vault entries (Reports / Notes / Knowledge)
 │         ├── Files (attachments)
 │         └── Tables (structured rows)
 ├── Agents (Orbits)
 │    ├── Events (run history)
 │    ├── Messages (your instructions + replies)
 │    └── Decisions (scoped directives)
 └── Org Chart (agents grouped by department)
```

## Trust attributes on every Vault entry

Each Vault entry carries:

* **Trust level** — `exploratory`, `conclusion`, or `authoritative`. How confident the content is. See [Trust Levels](/concepts/trust-levels).
* **Source trust** — how reliable the *source* (agent or operator) is, assigned automatically.
* **Provenance** — which agent filed it, when, and with what reasoning.
* **Version history** — prior versions are archived, never deleted.

## Identity and versioning

* Agents are identified by a stable `agent.id`. Reuse it across runs.
* Files and tables version in place via a stable `ref` / `table_ref` — re-uploading under the same ref archives the old version and shows the new one.
* Decisions move through states: open → done (or closed).

<Card title="The tools that read and write this" icon="rectangle-terminal" href="/reference/mcp-tools">
  See how each object is created and queried.
</Card>