Skip to main content
A workspace is the container for everything Hyper knows about your team. All shared documents, captured memories, and team-member context live inside a workspace. Every agent that connects to Hyper connects to a specific workspace — that’s what determines what context they receive and where their memories are written.

​
Personal vs. team workspaces

Hyper supports two types of workspaces, and you can have both at once:

Personal Workspace

A private memory space for individual use. Only you can read or write to it. Use it for personal project context, draft notes, or trying Hyper before involving your team.

Team Workspace

A shared memory space for your whole team. All members can query it, and writes from any connected agent are visible to everyone. This is where your company brain lives.
You create a workspace during the desktop app’s onboarding flow. There’s no limit on the number of workspaces you can belong to — many teams run one workspace per product or environment.

​
Creating a workspace

1

Open the desktop app

Launch Hyper and navigate to the Workspaces section from the sidebar. Click New Workspace.
2

Name your workspace

Give it a clear, recognizable name — usually your company name, team name, or project name. This name appears in agent briefings.
3

Choose a type

Select Personal if this is just for you, or Team if you plan to invite colleagues. You can’t change this after creation, so pick intentionally.
4

Complete setup

Hyper initializes the three core org documents — org/identity.md, org/decisions.md, and org/goals.md — and creates your personal folder at people/{your-name}/. You’re set as the workspace admin automatically.

​
Joining a workspace

If a teammate has already set up a workspace, you can join it in two ways:

​
Roles and permissions

Every workspace member has one of two roles:
Admins have full read and write access to the entire workspace, including:
  • org/identity.md — the organization’s foundational description
  • org/decisions.md — the log of team decisions
  • org/goals.md — current priorities and sprint targets
  • All people/{name}/* documents for every member
  • Workspace settings: invite links, integrations, billing, member management
When Hyper’s memory pipeline writes to org-level documents, it uses admin-level permissions. Only humans with the admin role can make direct edits to org docs outside of the AI pipeline.
Members can read all workspace documents and write to their own area:
  • people/{their-name}/* — their personal notes, status, and context
  • Their assigned project or feature areas (if scoped by an admin)
  • feed.md — the shared team activity feed
Members cannot edit org-level documents directly. Their remember() calls that target org-level content are routed through the same LLM synthesis pipeline as admins, but the final write requires no manual approval — Hyper handles it.
Be careful about who you grant admin access to. Admins can modify the org-level documents that shape every agent’s understanding of your company. A poorly edited org/identity.md will propagate misinformation to every future session until corrected.

​
Switching workspaces

The active workspace is determined by whichever workspace you have selected in the Hyper desktop app. When you call connect() in your AI client, Hyper reads the active workspace from your desktop session automatically — no extra configuration needed. To switch workspaces:
  1. Click the workspace name in the Hyper desktop app’s top navigation.
  2. Select the workspace you want to switch to.
  3. Start a new AI session and call connect() — it will brief you for the new workspace.
Workspace switching happens at the desktop app level, not the MCP level. If you want an agent to connect to a specific workspace regardless of your desktop selection, pass the join_token parameter directly to connect().

​
The members page

Every team workspace includes a Members view in the desktop app. This page shows each team member alongside an AI-generated summary of what they’re currently working on, derived from their recent remember() calls and people/{name}/* documents. These summaries update automatically as agents record new context. They’re designed to answer the question “who knows the most about X right now?” without anyone having to write a manual status update.
Encourage every team member to call remember() with meaningful context at the end of each session. The members page is only as useful as the memory your team puts into it.