Agent Team overview

What is Agent Team?

A team member is a named helper, like QA or Dev, that you connect from an AI tool such as Cursor, Claude, ChatGPT or Grok. Every connection gets its own key, so you can see who saved what, narrow what each member can use, and cut one off without affecting the others.

• All your team members share the same memory and task list. Access decides what each one can use.
• Each member also has its own private memory that only it (and you) can see.
• Team is in the My Docs sidebar at /my-docs/team/.

Agent Team is rolling out gradually, so you may not see it on your account yet.

Create a team member

Choose Create a team member and fill in:

• Name, for example "QA".
• Handle, for example qa. This is how tasks are assigned to the member, and it cannot be changed later.
• Label (optional), a short line about what the member does.
• Access, set per area:
  • Memory: off, read only, or read and write
  • Tasks: on or off
  • Projects: off, read only, or read and write
  • Skills: off, read only, or read and write
  • Documents: on or off

You can change access later from the member's page.

Connect a tool

Open a member and choose Connect a tool. Pick the tool, then copy the setup it shows you.

• The key is shown once. Copy it right away. It cannot be shown again.
• A member can have several connections, for example Cursor on your laptop and Claude on your desktop. Each has its own key.
• You can replace a connection's key or revoke it without touching the others.

A new member shows "Waiting for first use" until its tool makes its first successful call. Connecting alone is not enough.

Auto-created members

When you connect a tool the usual way, Gryz creates a member for it automatically, tagged Auto-created. You can rename it. Its access stays full and cannot be narrowed.

To limit a tool, create a team member for it instead and connect the tool with that member's key.

Private memory

Each member can save memories that only it can read.

• A member sees its own private memory and the shared project memory it has access to. It never sees your personal memory.
• Other members cannot see a member's private memory, and cannot tell that it exists.
• You can read, add and delete a member's private memory from its page.
• Private memory counts toward your total memory limit. There is no separate limit per member.

From an AI tool, a member saves to it with remember using scope: "agent", and reads it back with recall. See Agent Memory.

See what a member is doing

• The Team page lists every member with its status, access, connections and when it was last active.
• A member's page shows its connections, recent activity and a history of changes such as renames, access changes, pauses and key changes.
• A member can ask for its own profile with the get_profile tool: who it is, how much private memory it has, what is open for it, and its recent activity. You can also ask for the team summary, which lists the members and how many seats you have used out of your plan's limit. See the MCP tool reference.

Pause, resume or remove a member

• Pause stops the member from working until you resume it. Its keys stay in place.
• Remove ends the member and revokes its keys. Removed members stay in a collapsed Removed list so the history is kept.

Plan limits

• Auto-created members and removed members do not count toward the member limit.
• If you downgrade, your existing members keep working. You just cannot add new ones until you are under the limit.

See plan limits for the full comparison.