Recon: The tmux Dashboard That Finally Makes Multi-Agent Claude Code Workflows Manageable

What It Does — A Visual Command Center for Your Claude Code Agents

Recon is a terminal user interface (TUI) built specifically for developers running multiple Claude Code sessions in tmux. Instead of juggling terminal tabs or tmux windows, Recon gives you a unified dashboard showing every active Claude Code agent, their status, context usage, and what they're working on—all from a single keybinding.

The dashboard displays each agent as a pixel-art creature in a room (grouped by working directory), with animations indicating their status:

• Working/Input creatures animate — actively streaming or waiting for your approval
• Idle/New creatures stay still — done or haven't started yet
• Color-coded context bars — green/yellow/red showing token usage

How It Works — Deep tmux Integration Without Heuristics

Recon doesn't guess or parse processes. It reads directly from Claude Code's own session files and tmux's internal state:

# Recon reads from:
~/.claude/sessions/{PID}.json  # Claude Code's session tracking
~/.claude/projects/…/*.jsonl   # Project context
# Plus tmux's list-panes and capture-pane for real-time status

Each Claude Code instance runs in its own tmux session, and Recon monitors them all through the tmux server. The status detection inspects the Claude Code TUI status bar text at the bottom of each pane, giving you accurate real-time information without polling or guessing.

Setup — Install and Start Managing in Minutes

# Clone and install
git clone https://github.com/gavraz/recon
cd recon
cargo install --path .

![Tamagotchi View](https://github.com/gavraz/recon/raw/main/assets/view.png)


# Requirements: tmux and Claude Code already running
# Launch Recon from any terminal
recon

Once installed, Recon automatically discovers all your running Claude Code tmux sessions. The interface uses vim-like navigation:

• j/k to navigate between sessions
• Enter to switch to a session
• v to view session details
• r to refresh
• q to quit

When To Use It — Multi-Agent Workflows Made Practical

Recon shines when you're running parallel Claude Code sessions for different tasks—exactly the workflow Claude Code encourages. Here are specific use cases:

1. Code Review While Developing
Keep one agent reviewing a PR (webapp::pr-452) while another refactors your API (myapp::feat/auth). Recon's highlight shows which needs your attention.

2. Debugging Complex Systems
Run separate agents on different services (api-refactor, debug-pipeline, write-tests) and monitor their progress without context switching.

3. Long-Running Tasks
Start a heavy analysis in one session (using Opus 4.6 with 1M context), then switch to other work. Recon's "Last Active" column tells you when it's done.

4. Side Monitor Workflow
The pixel-art visualization is designed for glanceability. Keep Recon on a secondary monitor while you code—you'll instantly see when an agent finishes or gets stuck.

Pro Tips for Maximum Efficiency

1. Bind Recon to a tmux prefix key
   Add to your ~/.tmux.conf:

   bind-key R run-shell "recon"
   

   Now prefix + R brings up your dashboard from any tmux session.

2. Use the room grouping
   Agents are automatically grouped by working directory in 2×2 grids. Zoom into a room with 1-4 to focus on related tasks.

3. Watch the context bars
   Green/yellow/red coloring shows token usage at a glance. Red means an agent is approaching context limits—time to summarize or start fresh.

4. Kill idle sessions
   Recon makes it obvious which sessions are done. Clean them up directly from the interface to free system resources.

Why This Changes Multi-Agent Development

Before Recon, managing multiple Claude Code sessions meant:

• Remembering which tmux window had which task
• Manually checking each session's status
• Wasting time switching between contexts

Now you have a single-pane-of-glass view of your entire Claude Code ecosystem. The pixel-art visualization isn't just cute—it provides immediate status recognition that text alone can't match. When a creature animates, you know something's happening. When it's still, you can ignore it.

This is the missing piece for developers who want to use Claude Code's multi-agent capabilities seriously. Instead of limiting yourself to one or two sessions because they're hard to manage, you can run as many as your system handles—and actually keep track of them all.