We’re building Ever Works in public. Some things may be incomplete, missing, or broken while we continue improving the platform.We’re building Ever Works in public — expect a few rough edges.
Feature · Your team
In Ever Works your team is people and Agents working side by side. Organizations hold the work, invitations bring people in, Work roles decide who may change what, Teams describe who owns what, and every Agent carries its own scope, budget and permissions. It is the structure that decides who, human or AI, is allowed to do what.
Hold as many as you need in one account and switch between them from the top of the sidebar. Missions, Ideas, Works, Agents, Teams and Memory all belong to the Organization you pick.
Settings → Organization → Members takes an email address. Links are single-use, bound to that address, valid for seven days by default and can be revoked while pending. The invite form also tells you up front that someone you invite can see every Organization in your account.
Owner, Manager, Editor and Viewer, set per Work. Managers invite people and change roles, Editors edit content and trigger generation, Viewers read, and only the Owner can delete the Work.
Each Agent has a name, a title and a scope: the whole account, one Mission, one Idea or one Work. It also gets its own model, heartbeat and budget, and its instructions live in files you edit on its Instructions tab.
Put Agents and people on one roster as Lead or Member, nest teams up to ten levels deep, name a manager Agent, and attach the Works and Agents a team owns. The Org Chart draws humans and Agents together, and the Agents Chart shows the reporting lines on their own.
Eight permission flags (create agents, assign tasks, edit skills, edit instructions, spend budget, commit to the repository, open pull requests, call external tools) are all off on a new Agent. A tool that is off is never offered to the model.
With Require approval, every side-effectful action an Agent proposes queues for you with its risk flags spelled out, and you approve or reject it from the dashboard or the Inbox. Autonomous mode lets unflagged actions through, and flagged ones still wait.
Pause or resume any Agent from its header, and an Agent that fails three runs in a row pauses itself. Archived Agents stay on the Archived tab and come back paused when you restore them.
From the workspace switcher, choose Create Organization. Start blank, or pick one of more than thirty prebuilt companies, whose Agents are imported paused so you can review them first.
Send invitations from Settings → Organization → Members, then give each person a role on the Works they should touch.
Teams → New Team: choose a parent team and a manager Agent, then add Agents and people to the roster as Lead or Member.
Keep new Agents on Require approval with a small budget, watch a few runs, then loosen one layer at a time.
Some parts of this model are simpler than they look, and it is better to know that before you rely on them.