FridayOS 1.0: an Obsidian-based AI second brain with a six-region blueprint

What FridayOS 1.0 actually is, and who it is written for

FridayOS 1.0 is the renamed FridayOS-Lite, an open-source practice project from Robinson AI Lab. The README describes the lab as a public-interest AI research community working on AI management concepts and practical experiments, and calls FridayOS its flagship initiative for connecting human knowledge, experience and action with AI. The repository is Python-flavoured at the top level, but the shipped artefact is not a Python library: it is a set of Markdown guides, a blueprint file, a demo vault and a generator script. The README states plainly that this version focuses on personal knowledge management, and that a Feishu bot, automated message capture and a complete set of operational safeguards are outside its current scope. Broader task collaboration and Agent OS capabilities are described as ongoing research directions.

The intended reader is stated without hedging: people curious about AI who want help remembering, organizing and finding information, and who are willing to follow a guide and share what they learn. Prior software-development experience is not required. That is an unusual audience for a repository that asks you to install Node.js, an npm-installed coding agent, a provider-switching tool and an Obsidian community plugin. The guides are the bridge. If you already run a note-taking setup you are happy with, the value here is the structure and the prompt discipline, not the tooling.

The six-region brain and the folder-is-your-brain contract

The central mechanism is a fixed folder taxonomy. FRIDAY-BLUEPRINT.md defines six regions: inbox/ for fast capture and later sorting, exec/ for three to seven things in motion plus weekly and daily plans, wiki/ for material worth keeping and wikilinked, skills/ for repeated work turned into routines, raw/ for source material kept read-only, and system/ which holds the brain's contract in a file named CLAUDE.md. Notes stay as local plain text, which the README frames as never being locked in.

The claim that this layout is what makes Friday cheap to run is the most interesting design argument in the README. Aggregate queries are said to scan structured frontmatter rather than whole documents, every domain is said to have an overview map so a query starts from one page instead of eleven, questions are said to open only the region they belong to, and routines live in skills/ once and are invoked by name instead of being re-explained. Those are plausible token-economy choices, but the README also concedes that model costs depend on the provider, model, context size and usage, and tells readers to evaluate costs from their own usage records. Treat the frugality claim as an architecture description, not a measured result. Nothing in the repository as described publishes a token count or a cost comparison.

Installing the stack and building the brain in one sentence

INSTALL.md is the setup document, and the README orders readers through TOOLS.md first, then INSTALL.md, then FAQ.md if stuck, with an explicit instruction not to skip. The stack is Node.js as the foundation, Obsidian as the notes application, Claude Code installed through npm (the README points at a China mirror), cc-switch from ccswitch.io to wire the engine to a cheaper model, DeepSeek V4 from platform.deepseek.com as a no-VPN option, and Claudian from the Obsidian community plugin gallery to bring the AI into the sidebar. The README gives no version pins, so install the current releases of each.

The README does not print an install command for Claude Code; it names npm and a China mirror and leaves the specifics to INSTALL.md. What the README does give is the sentence that builds the brain. Download FRIDAY-BLUEPRINT.md, put it where the agent can read it, and tell Friday to build from it:

build from this

According to the README, the six regions then appear by themselves. What you should see afterwards is a vault containing inbox/, exec/, wiki/, skills/, raw/ and system/, with CLAUDE.md inside system/. If the folders do not appear, the FAQ.md troubleshooting page is the documented next stop. There is no published command for installing the plugin either: Claudian is enabled from Obsidian's community plugin gallery.

The demo brain is the fastest way to judge the structure

An empty vault tells you nothing about whether a six-region layout works, and the project acknowledges this directly. It ships 示例大脑-云栈科技.zip, a sandbox built around a fictional 100-person IT company with employee files, projects and customers, all wikilinked across the six regions, plus 11 demo prompts arranged in three acts for exploring retrieval and cross-file analysis. The workflow is to download the zip, unzip it, open the result as a vault and follow the demo manual inside.

One constraint matters before you spend time on it: the README states the demo content is Chinese-only. If you do not read Chinese, the sandbox still shows you the folder structure and the wikilink density, but the 11 prompts and the fictional company data will not demonstrate retrieval quality to you. That is a real gap for an English-language reader evaluating whether the structure suits their own notes. The repository does carry English versions of some documents, including README.md, LAB.md and the blueprint, but the demo vault itself is not translated according to the README.

Vibe Knowledge Management, and where the claim is untested

The README names its working style Vibe Knowledge Management, by analogy with vibe coding: you drive the knowledge base in natural language and the AI does the managing. The promise is that you never create folders, tag notes or fix links by hand, and that filing, linking and retrieval become Friday's job. VKM is described as the core of Friday, with FridayOS 1.0 as its smallest complete form.

The project is unusually candid about the limits of that claim. The README says this is a new way of working, that how well it replicates beyond the original team and how quickly newcomers internalize it still need more users to verify, and that this repository exists precisely to gather that evidence. That is a fair statement of an open question, and it should shape how you read the rest of the material. The six-region taxonomy is a hypothesis about how personal knowledge should be arranged for an AI to traverse cheaply. The demo brain tests it against synthetic data. Nothing described here tests it against a real, messy, years-old note collection, which is the case most readers actually have.

Where FridayOS 1.0 is the wrong tool

The scope statement is the clearest limitation, and it is worth taking literally. A Feishu bot, automated message capture and a complete set of operational safeguards are outside this version. If your problem is getting messages out of a chat platform and into a searchable store without manual steps, this project does not solve it today. A file named 飞书_FridayOS-Lite.md sits at the repository root, which suggests Feishu integration has been explored, but the README does not describe it as a shipped capability and no release notes were retrieved to confirm otherwise.

There is a second boundary. The README distinguishes the current version from an Agent OS, which it defines as a collaboration environment connecting AI agents, knowledge, tools and tasks, and places broader task collaboration in ongoing research. So if you are looking for multi-agent orchestration, tool calling or task routing, you are reading a roadmap, not a product. The system/ region holds CLAUDE.md as the brain's contract, which is the closest thing to an agent configuration surface in this version, and it governs one assistant over one vault. The installation path is also a chain of third-party dependencies: Node.js, Obsidian, Claude Code, cc-switch, DeepSeek and a community plugin. Each link can break independently, and the README routes troubleshooting to FAQ.md rather than offering a fallback stack.

How it compares with plain Obsidian and with agent frameworks

Against plain Obsidian, the difference is not the editor but the contract. Obsidian gives you folders, tags, backlinks and a plugin ecosystem, and leaves the organizing method to you. FridayOS 1.0 fixes the method: six named regions with defined purposes, an overview map per domain, read-only raw material, and a CLAUDE.md file that acts as the assistant's operating instructions. You are trading flexibility for a structure an agent can traverse predictably. If you already have a working PARA or Zettelkasten setup, the six regions will overlap with it and you will be re-filing notes to fit.

Against general agent frameworks, the difference runs the other way. Those projects expose orchestration primitives and expect you to write the agent. FridayOS 1.0 ships no runtime at all in the sense of code you import; it ships documents, a demo vault and a generator script, and it delegates execution to Claude Code running inside Obsidian. The trade is reach against setup cost. A framework gives you control over what the agent does; FridayOS gives you a pre-decided environment and a natural-language interface to it. Neither is a superset of the other, and the README's own framing of Agent OS as an aspiration rather than a delivered capability is consistent with that.

Maintenance, licence and what the repository does not promise

The repository is not archived, and the last push was on 2026-09-15, so the codebase is current as of this writing. That says nothing about release cadence: no releases were retrieved, so there is no version history to inspect and no upgrade path documented. Upgrades in practice mean pulling the repository again and re-reading INSTALL.md, because the moving parts are third-party tools rather than a package you pin. Obsidian, Claude Code, cc-switch and the Claudian plugin each update on their own schedule, and the README pins no versions for any of them. Budget for periodic re-reading of TOOLS.md and INSTALL.md rather than for dependency bumps in a lockfile.

The licence is MIT, which permits commercial and private use, modification and redistribution provided the copyright notice and permission notice are included. MIT offers no warranty and no patent grant, and it says nothing about the licences of the tools you install alongside it. Obsidian, Claude Code, cc-switch, DeepSeek and the Claudian plugin each carry their own terms, and model usage in particular is billed by the provider under those terms. That is a description of the licence text, not legal advice; check the terms of every component in your own context.

Editorial conclusion

Adopt FridayOS 1.0 if you want a guided, local-first way to run an AI over your own notes and you are comfortable installing Node.js, Obsidian, Claude Code, cc-switch and the Claudian plugin yourself. Do not adopt it if you need a Feishu bot, automated message capture or operational safeguards, all of which the README places outside this version's scope, or if you expect a task-collaboration Agent OS rather than a knowledge base. Before committing, read TOOLS.md and INSTALL.md end to end, confirm your model provider works through cc-switch, and decide whether the Chinese-only demo brain is enough to evaluate the six-region structure.

Frequently asked questions

Official sources

1. Issues
2. License: MIT
3. README
4. Robinson-AI-Lab/FridayOS1.0 on GitHub