Run an AI Engineering Team Without Hiring

Startup founders use amux to coordinate a fleet of Claude Code, Codex, or Gemini CLI agents that build features, fix bugs, write tests, and draft docs in parallel — managed from one dashboard or their phone. Updated July 2026.

In 2026, the constraint on a startup's engineering output is no longer team size — it is the quality of your task definitions and the discipline of your review process. amux is an open-source control plane for AI agent teams: it runs, monitors, and self-heals dozens of parallel coding sessions so a founder can ship what previously required a 5–10 person engineering team.

The limiting factor shifts from headcount to your vision. amux handles the coordination infrastructure: a shared kanban board, live session monitoring, a steering queue for mid-run guidance, token cost tracking, and mobile access so you can manage the whole operation from your phone.

What an AI Engineering Team Looks Like for a Founder

At any given time, a founder running amux might have sessions like this running in parallel:

All five run simultaneously. You see their statuses on one screen. When one finishes, you review the output, move the board card to Done, and queue the next task. The review bottleneck is you — not the agents.

From Idea to Shipped Feature in One Day

A typical amux workflow for a new feature:

1. Morning (15 min): Write clear task definitions for 3–5 agent sessions. Add them to the amux board. Register sessions in amux pointing at your repository.
2. Launch: Start all sessions simultaneously — amux start feature-payments bug-triage test-coverage. Each agent reads its CLAUDE.md spec and begins working independently.
3. Commute / meetings: Check session statuses from your phone. Send steering messages to redirect agents that go off-track. The board shows which tasks are in-progress.
4. Afternoon review (30–60 min): Agents have finished their tasks. Review diffs, accept or request revisions, merge clean PRs. Re-queue anything that needs iteration.
5. Overnight: Queue the next batch of tasks. Agents run while you sleep. Self-healing ensures they don't stall. Come back to completed pull requests in the morning.

Built-In Business Tools, Not Just Code

amux is designed for founders who need to run a business, not just ship code. Alongside the coding agent infrastructure, every amux installation includes:

• CRM — contacts with interaction logs, company, role, email, follow-up reminders. Log customer calls, track investor relationships, follow up on partnership conversations — all in the same dashboard where you manage your agents.
• Email (Gmail API) — send and reply to email directly from amux. Connected Gmail accounts get full threading, sent-from-your-domain delivery, and your real Gmail signature. Read inbox, search messages, compose replies without switching apps.
• Browser automation — a shared Playwright browser with saved auth profiles. Use it to automate repetitive web tasks: pulling reports, scraping competitor pricing, submitting forms, checking dashboard metrics.
• Notes — full Markdown documents accessible across all agent sessions via API. Your CLAUDE.md specs, product roadmaps, and meeting notes all live here.
• Scheduler — recurring jobs with cron expressions. Daily standups, weekly competitive analyses, nightly batch operations — define them once and they run automatically.

This is the difference between "a Claude Code wrapper" and an actual operating system for running your startup with AI.

Manage Everything from Your Phone

amux is the only AI agent dashboard built phone-first. Install the PWA on iPhone or Android, or download the native iOS app from the App Store. Every feature available on desktop is available on mobile:

• Live session status cards (working / waiting / idle)
• Peek at any agent's live terminal output
• Send steering instructions to redirect agents mid-run
• Manage the kanban board
• Check token spend across all sessions
• Read and reply to email, check CRM, add notes

For remote access from anywhere (not just your local network), set up Tailscale on the amux server and your phone. See the mobile management guide for full setup instructions.

amux vs. Hiring an Engineer vs. a Managed Service

The trade-off: amux agents need clear task definitions and human review of outputs. They are not autonomous junior engineers — they are powerful execution tools that amplify a founder's direction. The founders who get the most from amux are opinionated about what to build and disciplined about reviewing what the agents produce.

Open Source, Self-Hosted, No Per-Seat Fees

amux is MIT licensed. The entire server is a single Python file with zero external dependencies — no Docker, no complex build process, no cloud account required. You run it on your laptop or a small cloud VM. You pay nothing to amux. Your costs are:

• AI tokens — billed directly by Anthropic (Claude), OpenAI (Codex), or Google (Gemini CLI) at standard API rates
• Server hosting (optional) — a $5–10/month VM on any cloud provider if you want 24/7 agent operation. Not required if you run agents from your laptop.

There is no per-seat fee, no SaaS subscription, no vendor lock-in. Use the AI agent cost calculator to model your monthly spend based on parallelism, model mix, and hours per day.

How to Get Started as a Founder

1. Install amux (5 minutes):

   git clone https://github.com/mixpeek/amux && cd amux && ./install.sh

2. Register your project:

   amux register myapp --dir ~/Dev/myapp --yolo

   The --yolo flag lets Claude Code run without per-command approval — appropriate for trusted codebases where you review at the PR level, not the command level.

3. Start the dashboard:

   amux serve  # → https://localhost:8822

4. Write a CLAUDE.md in your project root. This is the context file every agent reads on startup — project architecture, coding conventions, what to avoid, how to run tests. Clear instructions here are the single highest-leverage thing you can do to improve agent output quality.
5. Launch your first agents from the dashboard or CLI. Add tasks to the board. Watch them work.
6. Install the mobile app so you can monitor and steer from anywhere.

Want Someone to Set It Up for You?

If you want the benefits of an AI engineering team without spending time on setup, amux Concierge is a managed onboarding service. The team installs and configures your agent fleet, sets up your first automations, and hands it off to you running. It is the fastest path from "I want this" to "it's working."

git clone https://github.com/mixpeek/amux && cd amux && ./install.sh
amux register myapp --dir ~/Dev/myapp --yolo
amux serve  # → https://localhost:8822

Frequently Asked Questions

Can a non-technical founder use amux?

amux requires basic technical comfort — you run it from the command line and configure it with a text file. If you can use Git and a terminal, you can run amux. For founders who want the benefits without the setup, amux Concierge provides managed onboarding. The ongoing operation (sending tasks to agents, reviewing output, managing the board) is accessible from the amux dashboard with no command-line required.

How much does it cost to run an AI engineering team with amux?

amux itself is free and open source (MIT license). You self-host it on a $5–10/month cloud VM. Your costs come from AI tokens: Claude Sonnet 4.6 costs roughly $3 per million input tokens and $15 per million output tokens. A typical 8-hour coding agent session costs about $20–50 in API costs. Running 5 agents in parallel for a full day is therefore around $100–250 — a fraction of a single engineer-day. Use the cost calculator to model your specific scenario.

What's the difference between amux and just running Claude Code?

Claude Code runs one session at a time in one terminal. amux runs dozens simultaneously — with a shared kanban board, live status monitoring for all sessions, a steering queue, cost tracking, CRM, email, browser automation, and mobile access. amux does not modify Claude Code; it manages tmux sessions that run it headlessly. The difference is a single contractor working alone versus a coordinated team working from a shared dashboard.

Does amux work with Codex and Gemini CLI, or only Claude Code?

amux works with any coding agent that runs in a terminal: Claude Code, OpenAI Codex CLI, Antigravity CLI (formerly Gemini CLI), or any other tool. You can run a mixed fleet — a Claude Code agent for complex reasoning, a Codex agent for fast code generation — and compare output quality and cost across providers.

What happens when an agent crashes overnight?

amux's self-healing watchdog detects the crash within seconds, restarts the session, and replays the last message so the agent picks up where it left off. Context overflow triggers automatic compaction. You can start a multi-hour run at midnight and come back to completed work in the morning, reviewed from your phone.

Can my co-founder or contractors see what the agents are doing?

Yes. The amux dashboard is a web app — any browser on your local network (or via Tailscale for remote access) can view session statuses, peek at agent output, and update the shared kanban board. There is no per-seat pricing. You can give a contractor read access to specific sessions without exposing your full codebase or SSH credentials.

Related Resources

• How to Run an AI Engineering Team (2026 Guide)
• Run Your Agents from Your Phone
• Running AI Agents Overnight Unattended
• amux for Solopreneurs
• amux Concierge — Managed Setup
• AI Agent Cost Calculator
• amux vs. Devin