PlumHelp
Help · Workspace settings

AI in Plum

Plum uses Claude, on your own Anthropic account, to answer questions about your workspace and to draft suggestions: deal lines from notes, next steps, renewal follow-ups and delivery briefs. Nothing changes until a person accepts it.

Owners and admins connect the key and choose the model. Anyone who works accounts can ask and accept suggestions.

Connect your Anthropic key

  1. In the Anthropic Console, create an API key.
  2. In Plum, open Workspace → Integrations and go to AI.
  3. Paste the key and select Save key. Plum checks it with Anthropic before saving it.

The key is stored encrypted with your workspace's own key and can't be read back by anyone, including owners. Usage is billed to, and limited by, that Anthropic account; Plum adds no limit of its own. Remove the key at any time to turn AI off.

Choose a model

Claude Opus 5.5
The most capable, and the default.
Claude Sonnet 5.5
Faster, at about half the cost.
Claude Haiku 5.5
The fastest and cheapest, for simple drafting.

What Claude sees

Only what the person asking can see in Plum. Everything sent to Claude is gathered as that person, under the same rules that decide what they can open, and only what the question needs. Meeting transcripts are included only if your workspace keeps them. Anthropic doesn't train its models on what is sent through its API.

Suggestions, not changes

When Claude drafts something, such as the offers on a deal or a delivery brief, it appears as a suggestion with Accept and Reject. Accepting makes the change the same way you would by hand, so the change history names you. Owners and admins can see how many suggestions were accepted and rejected.

Claude can be wrong. Read a suggestion before accepting it, especially an amount or a date.

Related