Unpack it, run one init command, and a lab where AI researches AI spins up — whole, on your machine. 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 research. Unpack it and run one initialization command, and a self-contained AI company folder — your own instance — spins up. A five-stage research pipeline — search, verify, format, draft, critique — lets you run, on your own machine, the picture of an AI observing and verifying AI. 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 Research-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. This is the edition with its roles tuned for research. Here too, 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 research, right on your machine. The backbone is the same AI for AI idea. Rather than forcing an AI to operate software made for people, we design the tools assuming an AI will use them from day one. If MCP (a layer that connects an AI to external tools) is the connection layer, what this pack works on is a cognitive description layer, where the way an AI thinks is written down in language. At the center of this edition is a five-stage research pipeline: search → verify → format → draft → critique, each handled by a dedicated AI agent on top of S-ICL (a way to describe an AI's own way of thinking in language). It searches scholarly literature, verifies source reliability, formats citations, synthesizes a four-layer report, and finally reviews it critically through peer review — end to end. You can run an AI organization with a real division of labor for research, right on your own machine. What you get - Eight core agents plus a secretary, a research scout, and five research-focused agents: the organization's central agents, a secretary-equivalent that coordinates them, a research scout, and five agents specialized for research. - A five-stage research pipeline: search, verify, format, draft, and critique (literature search, source verification, citation formatting, report drafting, and peer critique) run in sequence, carrying out everything from literature search through reliability verification, citation formatting, four-layer report synthesis, and peer review, end to end. - A self-contained, one-init setup: unpack it and run one initialization command, and your own AI company folder is generated. 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 built in, with no dependence on any separate home base. - Eleven included documents: more generous than the outreach edition, including faq and purchaser_guide among its eleven docs — fully symmetric with the Japanese edition — so you won't get lost right after buying. Who it's for - People who want to run the whole flow — have an AI search, verify, organize, and critically review — as an organization with a real division of labor. - People who want to observe, on their own machine, the picture of an AI researching AI. - 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 v0.2.4-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 and improve 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, faq, purchaser_guide, and so on) and the research agent definitions are in English, fully symmetric with the Japanese edition in docs and pipeline. 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 edition also ships purchaser_guide and faq, so you won't get lost right after buying. - 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.