Unpack it, run one init command, and your own AI company for outreach spins up, ready to use. We don't take tools built for humans and make an AI use them. We design tools from the start on the premise that an AI will be the one using them. This pack puts that idea to work for a single purpose: outreach. Unpack it and run one initialization command, and a self-contained AI company folder — your own instance — spins up. Ten AI agents, running on an owner → secretary → departments chain of command, carry out the work when you simply ask in plain language. This is an early-access release (an alpha build for evaluation), and we'll tell you what's inside, plainly. What this product is The AItoAI Outreach-Focused AI Pack is a real product: it ships a language specification for running AI agents, together with finished products built on that spec, ready to use. It is not a collection of prompts. The clearest thing about it: unpack it, run a single initialization command, and a complete AI company folder — your own dedicated instance — is generated. A memory engine and a mechanism that improves the organization on its own are all built in, so it runs standalone without relying on any separate home base. Think of it as bringing up a small OS for outreach, right on your machine. The backbone is an idea we call AI for AI. Rather than forcing an AI to operate software designed for people, we design the tools assuming an AI will use them from day one. As MCP (a layer that connects an AI to external tools) spreads as a connection layer, what this pack works on sits one step before that: a cognitive description layer, where the way an AI thinks is itself written down in language. If MCP is the connection layer, this is the cognitive description layer. Concretely, it builds an outreach-focused set of AI agents on top of S-ICL (a way to describe an AI's own way of thinking in language). Just ask the secretary in plain language, and the organization handles column writing, social posting, monetization analysis, deliverable-spec drafting, and research, as a team with a real division of labor. What you get - An outreach-focused AI company of ten agents, running on an owner → secretary (CES) → departments chain of command. It comprises five outreach-tuned departments (column writing, social posting, monetization analysis, deliverable-spec drafting, and research) plus five operational-core agents (a secretary-equivalent, a system-architect-equivalent, a norms-and-motivation OS, knowledge search, and knowledge update). - A self-contained, one-init setup: unpack it and run one initialization command, and your own AI company folder is generated. It depends on no separate home base. - Memory and self-improvement built in: an independent OMC (an organizational memory core, so the AI organization holds its own memory), GA (a mechanism that improves the organization automatically), and an MCP server method are all bundled, so memory startup, connection, and organizational improvement all work from a standalone unpack. - Nine included documents: start_here, technical_setup, daily_operation, advanced_usage, architecture, harness_guide, reference, troubleshooting, and README, arranged so you can get going without getting lost. Who it's for - People who want to run an AI not as a one-off chat, but as an organization with a real division of labor. - People who want to see, on their own machine, what the AI for AI idea actually changes for a concrete purpose like outreach. - People who'd rather start by unpacking and running one init command, then observing, instead of wrestling with a complex build. Important notes (told plainly) - This is an early-access release. The version is v0.1.2-alpha, intended for evaluation and testing. Whether to put it into production is a decision we ask you to make yourself, after evaluating it. We are not going to overstate what it is. - The generation feature is not included. This pack is configured to operate, improve, and run outreach with already-shipped AI agents. It does not include a feature for automatically generating new AI agents. What you get is a finished set of AI agents, not the machinery that produces them. - There are technical prerequisites. Generating your dedicated folder uses Node.js, and running the agents uses Python 3.10 or later (on Windows, use py -3.11), along with some technical literacy such as running an initialization command. - On the English edition, honestly. The buyer-facing docs (start_here, troubleshooting, and so on) are in English. However, some internal AI-agent definitions and internal knowledge still ship with Japanese text; full English localization is a future task. We state this as a known alpha-stage limitation, without either hiding it or overstating it. - On verification strength. This build has known limitations and is still a product at the evaluation stage. Please try it on that understanding. After purchase and support - After purchase, you can download the full pack through BOOTH's digital product delivery. The download also includes a short "READ ME FIRST" PDF — please open it first. - Setup starts by unpacking and running a single initialization command. Prepare a Node.js environment (used to generate your dedicated folder) and a Python 3.10 or later environment, then follow the start_here and technical_setup docs inside the pack. - This is an early-access release for evaluation. We'll announce update policy and a contact point through AItoAI's posts on Medium / X. Individual customization of the implementation is out of support scope; we accept contact regarding defects in the product itself or discrepancies in the description. This product runs on tools that support MCP servers, such as ClaudeCode.