Projects overview
A project is where a team's work lives: a codebase (or set of repositories), the tasks against it, and the members who can access it. Projects give the agent scope — which repos it may read, which models apply, and how usage is attributed — and give your team a shared place to track progress.
What a project contains
Every project is a workspace with a consistent set of building blocks:
Create work items, move them through statuses, and save the views your team returns to.
Add people, set a project lead, and understand project vs org permissions.
Project lifecycle
Projects move through a simple lifecycle so admins can distinguish active work from archived history:
Health indicators (on track, at risk, off track, no update) surface on the project overview so leads and org admins can spot stalled work without opening every task.
Creating a project
New project
From the Projects app, choose New project. Give it a name, a short key (used in task IDs like
ATL-42), and optionally link one or more repositories. If your org requires project assignment before agent use, make sure the right people are added in the same session.Add members
Open the project's Members panel and add people from your organization. Assign project admin to whoever should manage repos, policy, and membership; everyone else can be a member. Designate a project lead — the person accountable for delivery.
Configure scope
Confirm linked repositories, default branch, and any project-level model or budget overrides set by your org admin. Project policy resolves after organization policy, so inherited defaults apply unless narrowed here.
Start working
Create tasks (or import from GitHub/Linear if integrated), save a default view for the team, and open the agent scoped to this project to begin making changes.
Working with the agent in a project
When you start an agent run from a project, 4RGE resolves effective policy for that scope:
organization policy → team policy (if any) → project policy → member overrides
That means the same member may use different models or repo access in two projects. Runs started from the desktop, CLI, or cloud agents all respect this chain and attribute usage to the project.
Example: kick off a focused fix from the CLI inside a linked repo:
cd ~/work/my-service
4rged agent run \
--project atl-platform \
--prompt "Fix flaky integration tests in pkg/auth" \
--output jsonThe run ID in the JSON output links back to the project's activity feed and org usage rollups.
Production app URLs
In production, project work lives at https://4rged.ai/projects and cloud agent runs start
from https://4rged.ai/agents. Both apps share the same Supabase-backed org boundary and
resolve effective policy from api.4rged.ai (/api/desktop/effective-policy) on session start.
Local development serves the Projects app on port 3009 and the Agents workspace on port
3004 — see Troubleshooting if API calls fail during setup.
Who sees what
Projects belong to your organization. Org admins can see all projects; members see the ones they're added to. Archiving a project does not delete tasks or audit history — it removes the project from default directories.
Larger orgs often create a planned project from a template: default views ("My work", "Blocked", "This cycle"), label set, and a standing integration checklist. Clone the pattern rather than reinventing structure for every initiative.