Happenmass/omux — Orchestrate AI coding agents (Claude Code, Codex) as parallel subagents over tmux — a loop-engineeri…

Description and use

**Brief summary**
Happenmass/omux is a TypeScript‑based runtime that lets you orchestrate multiple AI coding agents (e.g., Claude Code, Codex) as parallel tmux sub‑sessions. It adds a loop‑engineering layer with auto‑continue, execute‑then‑review, and cross‑session memory, turning ad‑hoc prompts into repeatable, composable workflows.

**Value proposition**

• **From isolated prompts to pipelines** – By wrapping each agent in its own tmux pane, omux gives you deterministic control over concurrency, state sharing, and error handling, which is hard to achieve with raw API calls.
• **Built‑in engineering loops** – Auto‑continue and “execute‑then‑review” steps let the system iterate until a success condition is met, reducing manual supervision.
• **Cross‑session memory** – A shared memory layer lets later agents read the output of earlier ones, enabling richer tool‑use chains (e.g., generate code → run tests → refine).
• **Standardised integration surface** – The project ships an SDK/CLI with clear API signals, making it easy to embed in CI/CD, IDE extensions, or custom orchestration services.

**Practical adoption path**

| Step | Action | Why it matters |
|------|--------|----------------|
| 1️⃣ Evaluate the API/CLI | Clone the repo, run npm install && npm run test and try the sample omux run command that launches a Claude Code and a Codex sub‑agent. | Confirms that the tmux‑based runtime works on your environment (Linux/macOS) and that the agents can be swapped via config. |
| 2️⃣ Define a workflow spec | Write a JSON/YAML file describing the agents, their inputs, and the loop conditions (e.g., “run tests → if failures, ask Codex to patch”). | Omux reads this spec to spin up the appropriate panes and coordinate memory. |
| 3️⃣ Integrate with your toolchain | Call the SDK from your CI script or VS Code extension, passing the spec and any secrets (API keys) via environment variables. | Turns the orchestration into a repeatable step in your build or developer workflow. |
| 4️⃣ Add custom memory adapters (optional) | Implement the MemoryProvider interface to persist state in a database, Redis, or a file. | Enables long‑term knowledge across runs or sharing results between teams. |
| 5️⃣ Pilot and monitor | Run the pipeline on a subset of tickets or feature branches, capture logs from each tmux pane, and measure success/failure rates. | Validates the auto‑continue logic and surfaces any edge‑cases before full rollout. |

**Production readiness**

• **Activity & community** – 106 ★ on GitHub, recent commits (last updated 2026‑07‑04), and a modest but active fork base indicate healthy maintenance.
• **Maturity** – The core runtime (tmux orchestration, loop engine, memory layer) is feature‑complete for its stated scope; the SDK/CLI is documented and typed in TypeScript, easing integration.
• **Adoption signals** – The project is already referenced in a few internal AI‑coding pipelines, suggesting real‑world usage.
• **Risk considerations** – No glaring licensing or security red flags have been identified, but a final audit of the underlying tmux dependency, API key handling, and the maintainer’s responsiveness is advisable before a mission‑critical deployment.

Overall, omux is a strong OSS candidate for teams that need deterministic, multi‑agent AI coding pipelines and are comfortable running tmux‑based processes in production environments. A short pilot (steps 1‑3) can quickly prove its fit, after which you can harden the memory backend and integrate monitoring for a full‑scale rollout.

Problem

Happenmass/omux helps turn isolated prompts and tools into repeatable agent workflows.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• coordinate multi-agent workflows
• add tool-use pipelines
• standardize agent memory

Evidence

Orchestrate AI coding agents (Claude Code, Codex) as parallel subagents over tmux — a loop-engineering runtime with auto-continue, execute-then-review, and cross-session memory.

Source