User guide#

This guide explains how to use every user-facing part of Remote: onboarding, projects, chats, five agent providers, scheduled tasks, workspace tools, previews, the shared Agent Browser, Git recovery, secrets, sharing, users, and server information.

Remote is built around one idea: create a project computer, give an agent broad authority inside that project, and keep the human in control through the browser.

A Remote chat beside its live project preview
An authentic product capture: one project chat, its agent controls, tool activity, and the running application in the same workspace.

Start here#

If this is your first session:

  1. Sign in and finish setup.
  2. Create a project and learn the sidebar.
  3. Create a chat and choose an agent.
  4. Send work, attach context, queue a follow-up, or answer an agent question.
  5. Open the result through Files, Terminal, or IDE.
  6. For a web app, use Preview and Inspect.
  7. For a real signed-in website, use the Agent Browser.
  8. Use Git History before or after risky changes.
  9. Use Scheduled tasks when a project chat should work again later without keeping the browser open.

Find a task#

I want to…Go to
Claim a new server or sign inFirst run and sign-in
Connect Claude, Codex, or KimiGlobal settings, users, and providers
Configure MiniMax for one projectGlobal settings, users, and providers
Sign in to Antigravity for one projectGlobal settings, users, and providers
Create, search, reorder, start, or stop projectsProjects and sidebar
Pick a provider, model, thinking level, speed, mode, or skillChat and agent controls
Attach files, queue work, cancel a run, fork, rewind, or mark unreadPrompts, context, and conversation
Browse or download filesFiles, Terminal, and IDE
Run a command manuallyFiles, Terminal, and IDE
Open the browser IDEFiles, Terminal, and IDE
Preview a local web app or select an elementPreviews and inspector
Let an agent use a website where I am signed inAgent Browser
Inspect commits, prepare a recovery point, or switch clean versionsGit history and restore
Schedule, pause, edit, run, or delete future agent workScheduled tasks
Add secrets, share a project, set limits, or recover a containerProject settings
Invite a user, change a role, choose a theme, or inspect the hostGlobal settings
Follow a complete end-to-end recipeEveryday workflows
Diagnose a problemTroubleshooting
Check whether a feature exists and who may use itComplete feature reference

The application surfaces#

flowchart LR Sidebar["Projects and chats"] --> Chat["Agent chat"] Chat --> Work["Files, Terminal, IDE, and Git"] Chat --> Schedules["Scheduled tasks"] Chat --> Web["Preview, inspector, and Agent Browser"] Sidebar --> Project["Project settings"] Sidebar --> Global["Global settings"]
SurfaceWhat it controls
Projects sidebarProject and chat creation, search, status, ordering, read state, forking, and deletion
ChatAgent selection, prompt context, streamed reasoning and tool activity, per-tab drafts and queues, questions, usage, and history
Workspace toolsThe durable /workspace through files, downloads, a shell, code-server, and Git
Scheduled tasksHost-owned one-time or recurring prompts that return to a project chat
Project settingsContainer state, diagnostics, resource limits, secrets, membership, and recovery
Global settingsAppearance, host-wide agent credentials, platform users, Google OAuth, and host metrics

Roles at a glance#

CapabilityServer adminProject member
Use a project and its chatsYesYes, when added
Create projects and chatsYesYes
Use files, terminal, preview, IDE, and browserYesYes, with the IDE caveat below
Create and manage scheduled tasksYes, including all ownersYes, for their own tasks
Read or change project secretsYesYes
Add or remove project membersYesYes
Start, stop, restart, inspect, or repair a projectYesYes
Change CPU, memory, or root-disk limitsYesNo
Delete a projectYesNo
Connect agent-provider accountsYesNo
Configure Google OAuth or global usersYesNo

IDE access caveat: the current IDE proxy checks that a person is a registered Remote user, but it does not enforce project membership. Treat every invited user as able to reach every project IDE until that boundary is hardened.

What persists#

The durable center is the project, not one chat or one container generation.

  • /workspace survives normal stop, restart, and container replacement.
  • Claude, Codex, MiniMax, Kimi, and Antigravity provider homes are separate durable mounts. Antigravity persists only /root/.gemini/antigravity-cli, not the entire .gemini tree.
  • The Agent Browser profile lives in the workspace, so site sessions can survive container replacement.
  • Chat metadata and event history live in the host control plane.
  • Scheduled-task definitions, claims, and run state live in the host control plane.
  • Packages or files installed elsewhere in a container root filesystem can disappear when the container is recycled.
  • Drafts and queued prompts are stored in sessionStorage per browser tab. They survive a reload in that tab, but not tab closure, a different tab, another browser, or another device.

For the architecture behind those guarantees, read Philosophy and Projects and containers.

Screenshot notes#

The screenshots in this guide are authentic captures of the application walkthrough recorded on July 22, 2026. They show demo projects and point-in-time UI states, not synthetic mockups. A screenshot proves the visible surface; architecture diagrams and source-backed reference pages document behavior that a still image cannot prove.

remote.futrx documentation