Skip to main content
Archive a branch you no longer work on to take it off your workers and out of your workspace limits without deleting it. Unarchive it when you need it again: it comes back as it was. Settings → Usage shows how much of your workspace limits and archive quota you use, project by project and branch by branch, and archives a branch from there too (Usage and limits).

What archiving does

The archive keeps the branch’s files and commits as r5d last synchronized them from your workers. It doesn’t wait for your workers, so it takes seconds. A worker synchronizes a change, a commit included, within seconds of making it; to be sure a change made a moment ago is included, synchronize first (r5dctl workspace sync). Whatever a worker changed in the branch and hasn’t synchronized is lost when it deletes the checkout: right away for a connected worker, when it reconnects for an offline one. The preview names the offline workers. A branch whose conversation is running, queued or waiting, or a sub-agent of one, can’t be archived while it runs. You are told which conversation to stop; an agent’s archive waits for it. The project’s primary branch is archived only with its project.

Archive and unarchive in the app

Archive branch… is in the menu (⋮) of the branch’s header, on the branch’s own pages, and on the branch page. A session whose branch is merged (its pull request merged, or its work is in the project’s primary branch) also shows Archive when you point at it in the sidebar. To archive several branches at once, check them on the project’s Branches page and select Archive. The dialog previews what archiving would do before you confirm; r5d never archives a branch by itself. Every page of an archived branch says so under its header, with Unarchive. The project’s Branches page lists its archived branches under Archived and hidden, with when they were archived, Unarchive and delete; Archived (/archived, from All archived work…) lists the archived work of all your projects, and suggests branches you might archive: merged ones first, then idle ones, with their pull request and the commits they have that aren’t in the primary branch.

Archive and unarchive with r5dctl

Preview what archiving would do, then archive:
Without --confirm, nothing changes: the preview lists each branch’s files and their size, its conversations, the schedules that would pause, the processes it would stop, the workers that are offline, and your archive quota. Find candidates and archived branches:
Read an archived branch’s files, which needs no worker (this works for live branches too: their files as r5d last synchronized them):
Unarchive it:
The branch comes back under its name. Every worker receives its files and sets up its checkout by itself, with the uncommitted changes it was archived with still uncommitted; dependencies such as node_modules/ are installed again. Unarchiving is refused when the files would take your workspace past its limits: archive or delete other branches first.

Archive a whole project

Archiving a project archives every branch of it, the primary one included. An archived project takes no new branch or conversation. Unarchiving it restores its primary branch and the branches archived with it; --only-primary restores the primary branch alone, and the others stay archived until you unarchive them. Unarchiving any branch of an archived project unarchives the project with its primary branch.

Delete archived work

r5dctl -p my-project delete branch old-experiment --confirm deletes an archived branch for good: its files, its history on r5d and its conversations. Deleting a project deletes its archived branches with it. Archived branches are kept until you delete them; nothing expires.

Agents

Agents archive and unarchive with r5dctl like you, with the same refusals; an agent can’t archive the branch its own conversation runs on. An agent that refers to an archived branch (@my-project/old-experiment) gets no checkout: r5dctl checkout path says the branch is archived and how to read it with files ls and files cat.