> ## Documentation Index
> Fetch the complete documentation index at: https://productbrain.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Core concepts
> Canonical vocabulary for the Chain, entries, prefixes, lifecycle, and surfaces.
**Early access** — Prefix sets and collection slugs evolve with the product; your workspace may still show legacy IDs (`BET-*`) while docs standardize on **work packages** language. [**Join the Product Brain Community**](https://community.productbrain.io) when terminology and UI disagree.
Read [The Chain](/docs/the-chain) for the story. This page is the **lookup** reference.
## Chain
The **Chain** is the workspace **source of truth**: typed entries, relations, history, and governance—shared by the app, `pb`, MCP, and (eventually) the public REST API.
## Workspace
A **workspace** is an isolated container: its Chain, members, keys, and settings do not leak into other tenants. Humans pick a workspace in the UI; agents select it implicitly through the API key.
## Entries
**Entries** are durable knowledge units. Every entry has:
* A **human-readable ID** with a type prefix (examples below).
* **Title / name** and rich **data** fields depending on collection.
* **Status** (lifecycle) and often **workflow** state for governed collections.
* **Collection** membership (defines schema and rules).
* **Relations** to other entries and a **version / history** trail.
## Entry types (prefix cheat sheet)
Prefixes map to collections in typical workspaces. Exact slugs can vary—use `pb collections list` or MCP `entries` when in doubt.
| Prefix | Common name | One-line purpose | Example title |
| -------------- | ---------------------------- | ------------------------------------ | ------------------------------------------------------------------------- |
| `DEC-` | Decision | A choice the team made | DEC-89 — Feature flags for user-visible work |
| `TEN-` | Tension | Unresolved pull or conflict | TEN-31 — Auth breaks on mobile |
| `STD-` | Standard | Constraint future work must respect | STD-12 — API changes require review |
| `BR-` | Business rule | Operational system rule | BR-150 — Enforce least privilege on keys |
| `INS-` | Insight | Learning from data, users, or events | INS-7 — 60% drop at step 3 |
| `BET-` / `WP-` | Work package | Shaped initiative for delivery | BET-42 — Onboarding redesign *(ID style depends on workspace generation)* |
| `FEAT-` | Feature | User-visible capability slice | FEAT-216 — MCP session tools |
| `ARCH-` | Architecture | Architectural commitment | ARCH-26 — Domain map |
| `PAT-` | Pattern | Reusable solution shape | PAT-61 — Flyout content from thinking layers |
| `FLO-` | Flow | Workflow / ritual | FLO-18 — Bridge Level 0→1 |
| `AUD-` | Audience / research artifact | Persona or research signal | AUD-1 — Enterprise CEO ICP |
| `OPP-` | Opportunity | Opportunity tree node | OPP-3 — Self-serve expansion |
| `KR-` | Key result | Measurable result | KR-12 — Reduce time-to-orient |
| `MIL-` | Milestone | Timeline milestone | MIL-4 — Public docs launch |
| `INI-` | Initiative | Program container | INI-1 — Hosted beta hardening |
> **Human vs agent:** you say “capture a tension about mobile auth.” **Your agent** chooses `capture` text and collection classification; you confirm before commit when required.
## Lifecycle (simplified)
1. **Draft** — safe to iterate; not treated as committed truth.
2. **Committed / active** — governed promotion succeeded; consumers should trust it for orientation.
3. **Deprecated / archived** — superseded or retired; still visible for history unless hidden by policy.
Workflow states (e.g. `shaped`, `building`, `shipped` for work packages) layer on top—see MCP `entries action=update` docs in the [generated reference](/docs/mcp/generated-reference).
## Relations (semantic graph)
Relations are **typed directed edges**. Examples:
* A **standard *governs*** a work package.
* An **insight *informs*** a decision.
* A **decision / work package *resolves*** a tension.
We do not enumerate every relation type here—use MCP `relations` and the generated reference for exhaustive lists.
## `thinkingLayer` and timelines (advanced)
Entries can carry **thinking layer** metadata (strategy vs execution vs governance signals) and participate in **timeline** views. UIs may still label these differently—ask your agent to `entries action=get` when you need ground truth.
## `doneWhen` and `buildContract` (work packages)
Shaped work packages often embed **done when** criteria and **build contracts** so humans and agents share acceptance tests. Field names live in entry `data`—inspect a real `BET-*` / `WP-*` entry in your workspace.
## Sessions
**Sessions** trace agent work: what was read, captured, or committed while a task ran. Start before writes, close after.
## Governance (pointer)
Governance mode (e.g. open vs consensus) changes whether commits create proposals. Deep dive: **coming soon** as a dedicated doc; until then see [FAQ](/docs/faq) and in-app settings.
## Surfaces
| Surface | Role |
| -------- | --------------------------------------------------------------------------- |
| Web app | Humans navigate Zen Shell, Bridge, Import, Brain Chat. |
| `pb` CLI | Terminal + automation. |
| MCP | Agent-native tools in Cursor, Claude, ChatGPT, Codex, … |
| HTTP API | Public REST surface **coming soon** for integrations that cannot speak MCP. |
