Gamma integration

It is the clearest example of a genuinely shared connector: it carries no entity data of its own, so every entity can use it without an isolation problem.

What it gives Plexward

Decks from brain content. An investment memo, a status report or a roadmap section becomes a deck without a copy-paste stage.
Exports to the formats a recipient expects.
A reusable template path, so an entity's decks look consistent.

Layer

account, and specifically shared. One Gamma connector per Claude account, used by every entity, declared as shared so the dashboard does not repeat it under each one.

{ "name": "Gamma", "layer": "shared", "svc": "Decks", "account": "all entities" }

Prerequisites

A Gamma account on a plan that permits generation through the connector.
The connector attached on claude.ai.

Setup steps

Attach the connector on claude.ai and authorize.
Confirm from the brain: claude mcp list | grep -i gamma.
Declare it once per entity with layer shared, or once globally, depending on how you want the dashboard to read.

Per-entity routing

## Shared Integrations
| MCP Tool | Service | Notes |
|---|---|---|
| Gamma | Decks | Shared across all entities. No entity data lives in Gamma by default. |

There is no NOT This Company entry for Gamma, and that is the point: a shared connector is one where using it from the wrong brain does no harm. Keep the shared list short, and be sure before you put something on it.

Verify

cd ~/brains/acme-brain && claude mcp list | grep -i gamma

Expected ✔ Connected. Then generate a two-slide throwaway deck and confirm it appears in your Gamma workspace.

Common failures

Generation is asynchronous. The call returns a job, not a deck. A skill that assumes a synchronous result gets an empty link.
Templates and regular decks are different objects. Only a template can be used as one, and the ids are not interchangeable.
The connector creates, it does not edit. Refinements happen in Gamma's own editor. Promising a user "I will adjust the deck" is a promise the connector cannot keep.
Plan limits on generation count surface as a failed job rather than a clear error.

What Plexward does with it once connected

Skills that produce a narrative document can offer a deck as a second output: the same content, generated once, not maintained in two places. Because the source stays in the brain, the deck is regenerable rather than a fork.

Gamma