Skip to content
Kadro

kadro / blog / product guide

Kadro ADE: A macOS Workspace for AI Coding Agents

Published7 min read

Kadro ADE is a macOS workspace for developers who use AI coding agents but do not want their work scattered across unrelated terminal windows, chat tabs, and half-remembered prompts. It keeps Claude Code, Codex CLI, Gemini CLI, raw shell panes, browser references, and multi-agent Teams visible in one local desktop app.

If you are searching for an AI coding agent workspace, the short answer is this: Kadro ADE is not another model and not another chat UI. It is the operating surface around the tools you already run in your terminal.

What Kadro ADE is

Kadro ADE is a native macOS desktop app built with Tauri, React, TypeScript, Vite, and xterm.js. The core unit is a workspace. Inside a workspace, you can create splittable terminal panes, launch supported agent CLIs, run local commands, keep test runners visible, and open a browser sidebar without leaving the app.

The product is designed around the way agentic development actually feels: long-running sessions, multiple providers, visible work, reviewable diffs, and a human operator deciding what gets merged.

The problem with one chat window

A single chat is fine for a question or a small function. It breaks down when the work has parallel tracks. A real feature may need one agent reading the codebase, another agent implementing, another reviewing the diff, and a terminal running tests the whole time.

In a normal chat workflow, those responsibilities collapse into one scrollback. You lose role boundaries, you lose the live terminal context, and you often review the work only after it has drifted. Kadro ADE gives that work a layout.

What you can run in one workspace

The current provider registry supports Claude Code, Codex CLI, Gemini CLI, and raw shell panes. Kadro launches AI CLI panes through your shell, probes provider binaries in Settings, shows install hints when a CLI is missing, and keeps a normal shell available for local commands.

  • Claude Code for Anthropic-powered coding sessions.
  • Codex CLI for OpenAI coding workflows.
  • Gemini CLI for Google Gemini sessions.
  • Shell panes for tests, dev servers, scripts, and repo work.
  • A browser sidebar for docs, downloads, find-in-page, and zoom.

For full-screen agent CLIs, Kadro uses a tui-replace rendering mode so terminal UIs stay readable inside panes. You still work with real PTYs, not a simulated terminal.

How Teams turn agents into a workflow

Teams are Kadro’s multi-agent layer. A Team starts from a mission brief, a repository, provider/model defaults, and a roster. The built-in roles are architect, builder, reviewer, and scout. Presets range from Quick Build with 3 agents to Hive Mind with 17 agents.

Each agent gets its own git worktree. The crew shares an append-only NDJSON notebook for durable mission state. Agents can coordinate through the MCP bridge, and the dashboard shows the agent rail, mission timeline, work area, Talk, per-agent terminal attach, and worktree diffs.

This matters because the operator can inspect actual work, not just ask an agent to summarize what it thinks it did.

Who Kadro ADE is for

Kadro ADE is built for developers who already use agent CLIs and want more control over the working environment. It is a fit when you want to keep multiple sessions visible, compare providers, run tests next to agent output, review isolated worktrees, or delegate a mission to a small crew instead of a single long prompt.

It is not trying to host models, replace your editor, or hide the filesystem. The app is local-first: it gives structure to terminal tools, git worktrees, and human review.

How to start

  1. Download Kadro ADE for macOS.
  2. Install Claude Code, Codex CLI, or Gemini CLI if needed.
  3. Open Settings and confirm provider status.
  4. Create a workspace with agent panes and shell panes.
  5. Start with a small Team before scaling to a larger preset.

If your current AI coding workflow is one busy chat thread and five terminal windows, Kadro ADE gives that work a cleaner shape: one workspace, visible panes, isolated agent work, and a dashboard built for review.