If you run multiple Claude Code or Codex terminals at the same time, you already know the pain: tabs everywhere, lost context, missed completions. I built Tide Commander to solve that.
It's a visual multi-agent orchestrator where your AI agents appear as 3D characters on a battlefield. Click to select, type to command, watch them work in real-time. It looks like a game, but internally it's a full developer toolkit.
Quick Start
bunx tide-commander
That's it. Node.js 18+, Linux or Mac, and Claude Code or Codex CLI in your PATH.
The Problem
AI coding agents work best in parallel — one on tests, one on features, one on bugs. But managing five terminals at once is chaos. Which one has context on the auth module? Did the test agent finish? Tide Commander puts everything in one visual interface with file diffs, a git-integrated file explorer, and real-time streaming output. For many workflows, an IDE becomes almost unnecessary.
Key Concepts
Boss Agents — Has context of subordinate agents, delegates tasks to the most capable one, summarizes worker progress. Talk to one Boss instead of juggling terminals.
Supervisor — God-mode observer. Auto-generates summaries when agents finish. You never miss a completed task.
Group Areas — Draw areas on the battlefield to organize agents by project. Assign folders to enable the file explorer.
Classes — Like COD or Minecraft classes: each has a 3D model, instructions (like claude.md), and skills. Create custom classes with your own GLB models.
Buildings — Functional 3D structures that control real infrastructure:
| Type | What It Does |
|---|---|
| Server | PM2 integration, real-time logs, CPU/memory stats |
| Database | MySQL/PostgreSQL/Oracle — SQL editor, schema browser |
| Docker | Container management, health checks, log streaming |
| Boss Building | Bulk controls for subordinate buildings |
Developer Tools
Despite the RTS aesthetic, this is developer-first:
-
File Explorer + Git Diffs — Side-by-side diff viewer, branch switching, merge conflict UI, vim-style navigation (
j/k/d/u/f/b/g/G//n/N) - Context Tracking — Mana bar per agent showing context window usage
- Snapshots — Save full conversations + modified files, restore later
-
Secrets — AES-256-GCM encrypted
{{PLACEHOLDER}}injection, never in logs - Permissions — Bypass or interactive mode, per agent
-
Spotlight (
Ctrl+K) — Fuzzy search for agents, files, areas, actions - Paste Anything — File paths auto-attach, screenshots paste inline, large text auto-compacts
- Commander View — All agent terminals in a grid, grouped by area
Three View Modes
Cycle with Alt+2: 3D (Three.js battlefield, 12 built-in models, custom GLB support), 2D (canvas, same features, lighter), Dashboard (status cards and metrics).
Skills System
Plugins with tool permissions and markdown instructions. Built-in: Git Captain, Notifications (browser/Android/desktop), Streaming Exec, inter-agent messaging, DB & server logs. Create custom skills and assign to agents or classes.
Architecture
User → React + Three.js → WebSocket → Node.js/Express → claude/codex CLI
Each agent is a real CLI process. Backend parses stdout events and streams to the frontend. Sessions persist across restarts. ~130k lines of TypeScript. All data stays local.
Also Included
Multiplayer (WSS), mobile + Android APK, custom hotkeys, Docker deployment, WSS debugger, HTML-rendered output (no terminal flicker), custom idle/working animations per class.
Free and Open Source
MIT license. No paid tiers. No sign-up. No telemetry.
npm i -g tide-commander@latest
tide-commander start
deivid11
/
tide-commander
Tide Commander (Claude Code agents Orchestrator UI)
🌊 Tide Commander
A visual multi-agent orchestrator for Claude Code and Codex
Tide Commander is a Claude Code and Codex orchestrator and manager that lets you deploy, control, and monitor multiple AI coding agents from a single visual interface. Spawn agents, assign tasks, and watch them work in real-time on an RTS-inspired 3D battlefield — or switch to a lightweight 2D canvas or a metrics dashboard.
🚀 Getting Started
Requirements:
- Node.js 18+
- Claude Code CLI (
claudecommand available in PATH) - OpenAI Codex CLI compatibility
Install and run:
# Run directly without installing (recommended)
bunx tide-commander
# Or install globally
npm i -g tide-commander@latest
tide-commander start
Command lifecycle:
# Start in background (default)
tide-commander start
# Stop the background server
tide-commander stop
# Check whether server is running
tide-commander status
# Show the latest server logs
tide-commander logs
# Follow logs in real time
tide-commander logs --follow
CLI flags (for…
Top comments (2)
The Boss Agent pattern is the part that interests me most here. We run parallel agents on client projects regularly and the biggest pain isn't spawning them, it's keeping them from stepping on each other's changes. One agent refactors an auth module while another is writing tests against the old interface. Does Tide Commander handle file-level conflict detection between agents, or is that still on the developer to manage?
Also, context window tracking per agent is underrated. We burned through so many tokens before we started monitoring that.
Hello Mahima,
Right now Tide Commander tracks which files each agent has modified, but it doesn't block concurrent writes to the same file automatically. Internally Claude Code and Codex detect if the file has been modified berfore doing an edit.
In practice, the Boss agent pattern helps a lot since the Boss knows what each subordinate is working on, it naturally avoids assigning overlapping files. The trick is creating a plan that is totally parallelizable. Boss have context of their assigned agents so can decide how to parellize before assinging tasks.
And yeah, context tracking was one of the first things I built. The mana bar per agent saves a lot of wasted tokens. You just know at a glance when to start a fresh session.