Skip to content

Getting started

Condr is a workbench for managing and orchestrating Agent command-line tools such as Claude Code and Codex in one place. Whether an Agent runs locally or on a remote server, Condr lets you monitor it and work with it from a single view.

Condr comes in two forms: the desktop app and the Headless Server.

Go to the download page to get the package for your platform, then unpack or install it and launch it.

The desktop app bundles the Server component and starts it with the app by default, so no extra setup is needed:

  1. Create a Workspace: click Open Project and choose a local project directory. Condr creates a Workspace and opens a terminal automatically.
  2. Install Agent integrations: go to Settings › Agent integrations and click Install next to each Agent you want supported, so that Condr can capture and track its state.
  3. Run an Agent: run the Agent command (for example claude) directly in the built-in terminal.

For cloud hosts, GUI-less dev machines or remote servers, only the core condr CLI needs to be installed:

Linux / macOS:

Terminal window
curl -fsSL https://condr.dev/install.sh | sh

Windows:

Terminal window
irm https://condr.dev/install.ps1 | iex

After installation, reopen your terminal window to pick up the updated environment variables, then start the service:

Terminal window
condr server start
  • Core concepts: learn the architecture model of Device, Server, Workspace, Pane and Agent.
  • Workspaces, Tabs and Panes: set up split-Pane layouts and give each Agent its own branch environment with Git Worktree.
  • Agent integrations: see the list of compatible Agents, lifecycle Hooks and the state-tracking definitions.
  • Remote connections: reach remote devices over SSH, TCP or P2P.
  • Agent automation: coordinate Agents with each other through the condr CLI.
  • CLI reference: the complete manual of commands and options.
  • Troubleshooting: common connection drops, state drift and how to diagnose them.