- The authoritative runtime owns Agent processes, session timelines, workspace files, Git, terminals, the browser, computer use, provider credentials, and durable state. It comes in three forms with the same core:
- The local runtime embedded in the desktop (the default).
- The headless runtime
vibex-server, running on your own server or always-on machine. - An SSH / WSL remote runtime, provisioned by the desktop on another machine or inside WSL2 (beta).
- Clients are the desktop workbench, the iOS / Android mobile apps, and the Vibex TUI. They render the runtime’s projections and send typed operations. Mobile and the TUI do not run Agents locally and do not keep a second copy of the workspace.
Start here
Install the desktop and complete your first Agent task.
Core concepts
Projects, workspaces, sessions, runtimes, and providers.
Vibex TUI
Use the same runtime from a terminal, over SSH, or in tmux.
Providers and gateway
Shared Providers, route groups, and the local model gateway.
Changelog
New features, fixes, and internal changes by release.
What you can do with Vibex
- Follow Markdown replies, reasoning, plans, tool calls, process details, attachments, approvals, and elicitation forms in a structured timeline.
- Compose input with
/commands and Prompts,@files and Agents,$Skills, and attachments. - Delegate work to child Agents, keep a task open across rounds, follow up, accept or cancel, and track long turns with goals.
- Let Vibex download, upgrade, and uninstall supported Agent runtimes, and configure model Providers per Agent.
- Fail over across shared Providers, route groups, and the local gateway, with protocol translation.
- Search, edit, save, rename, and preview files in the file tree and editor.
- Inspect Git status, diffs, history, and blame; create branches, commit, and push; run parallel tasks in isolated Git worktrees.
- Use a native PTY with ANSI emulation in multiple terminal tabs and splits.
- Watch pages in the embedded browser and let an Agent read and operate them; with approval, let an Agent operate the local desktop.
- Install MCP servers and Skills from in-app markets, write reusable Prompts, and scope each resource per Agent.
- Queue follow-ups, steer a running turn, schedule a send for later, or let auto-continue carry a turn to completion.
- Filter usage by time, Agent, Provider, model, project, and session, with explicit coverage states.
- Search every conversation, import local Agent history, and archive and restore sessions.
- Pair a phone with the desktop, then follow progress, answer questions, approve requests, and use the terminal remotely.
- Continue the same sessions in a terminal with the TUI, or through SSH / WSL / a self-hosted runtime.
Next
Quick start
Install the desktop and complete your first session.
Changelog
New capabilities, fixes, and internal changes by release.