dmux Workflows

Orchestrate parallel AI agent sessions using dmux, a tmux pane manager for agent harnesses.

When to Activate

• Running multiple agent sessions in parallel
• Coordinating work across Claude Code, Codex, and other harnesses
• Complex tasks that benefit from divide-and-conquer parallelism
• User says "run in parallel", "split this work", "use dmux", or "multi-agent"

What is dmux

dmux is a tmux-based orchestration tool that manages AI agent panes:

• Press n to create a new pane with a prompt
• Press m to merge pane output back to the main session
• Supports: Claude Code, Codex, OpenCode, Cline, Gemini, Qwen

Install: Install dmux from its repository after reviewing the package. See github.com/standardagents/dmux

Quick Start

bash
1# Start dmux session
2dmux
3
4# Create agent panes (press 'n' in dmux, then type prompt)
5# Pane 1: "Implement the auth middleware in src/auth/"
6# Pane 2: "Write tests for the user service"
7# Pane 3: "Update API documentation"
8
9# Each pane runs its own agent session
10# Press 'm' to merge results back

Workflow Patterns

Pattern 1: Research + Implement

Split research and implementation into parallel tracks:

Pane 1 (Research): "Research best practices for rate limiting in Node.js.
  Check current libraries, compare approaches, and write findings to
  /tmp/rate-limit-research.md"

Pane 2 (Implement): "Implement rate limiting middleware for our Express API.
  Start with a basic token bucket, we'll refine after research completes."

# After Pane 1 completes, merge findings into Pane 2's context

Pattern 2: Multi-File Feature

Parallelize work across independent files:

Pane 1: "Create the database schema and migrations for the billing feature"
Pane 2: "Build the billing API endpoints in src/api/billing/"
Pane 3: "Create the billing dashboard UI components"

# Merge all, then do integration in main pane

Pattern 3: Test + Fix Loop

Run tests in one pane, fix in another:

Pane 1 (Watcher): "Run the test suite in watch mode. When tests fail,
  summarize the failures."

Pane 2 (Fixer): "Fix failing tests based on the error output from pane 1"

Pattern 4: Cross-Harness

Use different AI tools for different tasks:

Pane 1 (Claude Code): "Review the security of the auth module"
Pane 2 (Codex): "Refactor the utility functions for performance"
Pane 3 (Claude Code): "Write E2E tests for the checkout flow"

Pattern 5: Code Review Pipeline

Parallel review perspectives:

Pane 1: "Review src/api/ for security vulnerabilities"
Pane 2: "Review src/api/ for performance issues"
Pane 3: "Review src/api/ for test coverage gaps"

# Merge all reviews into a single report

Best Practices

1. Independent tasks only. Don't parallelize tasks that depend on each other's output.
2. Clear boundaries. Each pane should work on distinct files or concerns.
3. Merge strategically. Review pane output before merging to avoid conflicts.
4. Use git worktrees. For file-conflict-prone work, use separate worktrees per pane.
5. Resource awareness. Each pane uses API tokens — keep total panes under 5-6.

Git Worktree Integration

For tasks that touch overlapping files:

bash
1# Create worktrees for isolation
2git worktree add -b feat/auth ../feature-auth HEAD
3git worktree add -b feat/billing ../feature-billing HEAD
4
5# Run agents in separate worktrees
6# Pane 1: cd ../feature-auth && claude
7# Pane 2: cd ../feature-billing && claude
8
9# Merge branches when done
10git merge feat/auth
11git merge feat/billing

Complementary Tools

ECC Helper

ECC now includes a helper for external tmux-pane orchestration with separate git worktrees:

bash
1node scripts/orchestrate-worktrees.js plan.json --execute

Example plan.json:

json
1{
2  "sessionName": "skill-audit",
3  "baseRef": "HEAD",
4  "launcherCommand": "codex exec --cwd {worktree_path} --task-file {task_file}",
5  "workers": [
6    { "name": "docs-a", "task": "Fix skills 1-4 and write handoff notes." },
7    { "name": "docs-b", "task": "Fix skills 5-8 and write handoff notes." }
8  ]
9}

The helper:

• Creates one branch-backed git worktree per worker
• Optionally overlays selected seedPaths from the main checkout into each worker worktree
• Writes per-worker task.md, handoff.md, and status.md files under .orchestration/<session>/
• Starts a tmux session with one pane per worker
• Launches each worker command in its own pane
• Leaves the main pane free for the orchestrator

Use seedPaths when workers need access to dirty or untracked local files that are not yet part of HEAD, such as local orchestration scripts, draft plans, or docs:

json
1{
2  "sessionName": "workflow-e2e",
3  "seedPaths": [
4    "scripts/orchestrate-worktrees.js",
5    "scripts/lib/tmux-worktree-orchestrator.js",
6    ".claude/plan/workflow-e2e-test.json"
7  ],
8  "launcherCommand": "bash {repo_root}/scripts/orchestrate-codex-worker.sh {task_file} {handoff_file} {status_file}",
9  "workers": [
10    { "name": "seed-check", "task": "Verify seeded files are present before starting work." }
11  ]
12}

Troubleshooting

• Pane not responding: Switch to the pane directly or inspect it with tmux capture-pane -pt <session>:0.<pane-index>.
• Merge conflicts: Use git worktrees to isolate file changes per pane.
• High token usage: Reduce number of parallel panes. Each pane is a full agent session.
• tmux not found: Install with brew install tmux (macOS) or apt install tmux (Linux).