markfulton/ai-employees

📚 Quick Overview

ai‑employees ships a set of ready‑to‑run AI agents that act like full‑time staff members for a small business. Each “employee” is a folder of plain‑text files that define:

• What role it plays (e.g., GTM Engineer, SEO, Web Dev, Social Media, Ad Manager, Sales, Customer‑Satisfaction, Chief of Staff).
• A schedule of 7‑8 recurring routines (daily briefs, drafts, data pulls, etc.).
• Guardrails that stop the agent from taking irreversible actions without your explicit click.

The repo provides the kits (the folder structure and sample prompts) and a tiny Node‑based installer (npx ai-employees hire …). After installing, the employee runs on the AI coding assistant you already use – Claude Code or any of a dozen compatible agents – driving your browser and PC just as you would, then delivering a concise morning brief.

---

🛠️ How It Works

1. Install – either via the web‑based “Agent Ops Club” wizard or manually with npx ai-employees hire <role> --to <folder>.
2. Agent runs – the chosen AI agent reads the INSTALL‑PROMPT.md, gathers information from your website, builds a dashboard, and registers the employee’s schedule on your machine.
3. Scheduled routines – at the times listed in SCHEDULE.md the agent executes a routine (e.g., generate an SEO article, draft a social post, check ad performance). Results are saved as plain‑text files; any publish, send, or spend step is held for your manual approval.
4. Morning brief – each day the employee posts a short markdown summary of what was done, what needs your attention, and any blockers.
5. Iterative improvement – on subsequent runs the agent re‑evaluates pages or data it previously touched, updating its own instructions automatically.

All logic lives locally; the only external calls are to the AI model you’re logged into (Claude Pro/Max, etc.). No credentials are stored, no background services are installed, and upgrades (npx ai‑employees upgrade) merge new kit files without overwriting any custom edits you made.

---

🚀 What You Get

Employee	Core Deliverables	Typical Routines
GTM Engineer	Launch positioning, outbound drafts, press forms, weekly scoreboard	8
SEO/AEO	Keyword research, daily article draft, publishing, indexing, rank review	8
Web Dev	Site health checks, error triage, small code changes, dependency review	8
Social Media	Drafted posts for each platform, veto window, reply drafts	7
Ad Manager	Account reads, creative sets, budget change list (approval required)	7
Sales	Prospect sweeps, first‑touch drafts, follow‑up sequencing	7
Customer Satisfaction	Inbox sweep, priority‑ranked replies, churn flags	8
Chief of Staff	Reads other employees’ logs, surfaces stalled work, suggests three next moves	7

All 60 routines are fully documented in the employees/ directory with example outputs.

---

💰 Cost

The agents themselves are free (MIT‑licensed). Running them consumes your existing Claude subscription usage – roughly 6 % of a Claude Max seat for the GTM Engineer’s daily jobs. No extra API keys or paid services are required.

---

🤝 Community & Support

• Agent Ops Club – free membership gives access to a walkthrough, live sessions, and a premium software library.
• GitHub – issues, discussions, and a CONTRIBUTING guide for new routine ideas.
• Guardrails & Security – detailed docs (GUARDRAILS.md, SECURITY.md) describe what the agents can and cannot do.

---

📦 Getting Started (One‑liner)

npx ai-employees hire gtm-engineer --to ./my‑gtm‑engineer
# then, in Claude Code (or any supported agent):
"Install the GTM Engineer from this folder."

The installer copies the kit, runs a self‑test, and tells you the exact prompt to paste into a fresh agent session.

---

📝 TL;DR

ai‑employees is a genuine open‑source project that lets you run autonomous, schedule‑driven AI “staff” on your own computer, using the same large‑language‑model assistant you already have. It’s focused on practical business automation, with strong safety guardrails and a clear, low‑friction install path.