Open specification · Public draft v1.1 · CC-BY-4.0
YAIBOS is an open spec for an AI business OS. The Base is the file-based system Samuel uses today: company context, agent skills, role agents and tools in a second brain you own. Business skills turn standard operating procedures (SOPs) into reusable instructions for your AI. Use a compatible harness, such as Claude Code or Codex, and switch without rewriting your business files.
Start with Base. Add an optional Cockpit to launch and follow skills, then Team features when shared work needs them. Each layer explains its benefits, dependencies and exit path.
├── AGENTS.md company context ├── SOP/ procedures ├── Tools.md MCP servers, CLIs └── Areas/ knowledge
These are implementation profiles, not release versions or pricing plans. Base is complete on its own. Cockpit and Team describe additional capabilities, with more dependence on the chosen integration.
L1 · The system Samuel uses today
Your second brain, skills, role agents and tools live in open files. Run them through an existing AI harness, with your own official subscription or API key.
Sync adapters, authentication and tool connections for each supported harness. No dedicated YAIBOS app required.
Change harness while keeping your business files. Verify tools, permissions and schedules in the destination.
L2 · Optional execution dashboard
Everything in Base, plus a catalog to launch skills, follow progress, handle approvals and read run history. Useful even for one person.
A dashboard integrated with its supported harnesses. Launch, tracking and cancellation depend on that integration.
Keep Base and export run records. You may need a different dashboard when you change harness.
L3 · Optional shared operations
Everything in Cockpit, plus shared context, enforced role permissions, area owners and a team audit trail.
Identity, shared storage, file sync and access enforcement. More infrastructure to maintain.
Export files, history and policies. Configure and verify access enforcement again in the destination.
The business files remain portable at every level. Identical interface features across harnesses are not guaranteed. Read the profile requirements →
Base connects your second brain, tools and skills through a supported harness. Cockpit and Team add optional capabilities on top of those files.
~/YAIBOS
AFTER SETUP AND VERIFICATION
AGENTS.mdsecond-brain/5 SOP/ (reusable agent instructions)the tools registry via MCPsecond-brain/
Open Markdown files in your second brain. Contains company guidelines, customer records, and AGENTS.md rules.
Model Context Protocol (MCP) servers, local scripts, billing APIs, CRM hooks, and sandboxed file operations.
Procedures written in Markdown (second-brain/5 SOP/) with readable goals, inputs, steps, approval rules and outcomes. Business skills put repeatable procedures into practice.
Forcing non-technical team members to craft raw prompts in consumer chat windows leads to silent failures and expensive mistakes.
A cockpit reads the same Markdown skills and adds launch controls, progress and approval handling through a supported harness integration.
These example roles require enforced file and tool access. Base instructions alone do not provide multi-user isolation.
Reads all areas. Exclusive approval on banking, strategy, and skill modifications.
/ (All folders)Runs Lead Qualification and CRM Enrichment skills. Writes qualified contacts to CRM only.
Projects/Leads/Runs Invoices Audit skill. Reconciles receipts against bank statements.
Areas/Finance/This simulation illustrates the optional Cockpit profile: browse skills, launch them and follow progress and outputs. Base already works through an existing harness. A real cockpit needs an integration with its supported harnesses; switching harness may change this interface.
POST /api/v1/sop/{id}/run · yaibos://run?sop={id} An implementation may add API or deep-link triggers for workflows shared between humans and AI. The example routes above are illustrative, not active endpoints. Availability depends on the cockpit and harness integration, and each trigger must respect the skill approval policy.
Extracts incoming inquiries, enriches company size and revenue via business registries, and updates CRM pipeline.
Extracts tables from incoming vendor PDF invoices, verifies tax IDs, and matches lines against monthly bank CSVs.
Ingests unassigned customer helpdesk tickets, categorizes priority, drafts contextual answers from product docs, and alerts on bugs.
Ready. Click [▶ RUN] on any skill card to simulate live loop.
Base keeps your context, skills, role agents and tools in open files. Cockpit adds a dashboard for execution and tracking. Team adds shared work and enforced permissions. The same six pillars apply across the three profiles.
Operations staff and assistants will not adopt a tool that starts with npm install.
Skills package reusable instructions, scripts and resources. Business skills make your SOPs executable by an agent.
Metered API keys make team bills unpredictable. Most people already pay for an AI plan.
Agents are only as good as the context they read, and company knowledge decays quietly.
An agent with blanket access to payroll, client data and email is one bad instruction away from a costly mistake.
Harnesses and models change every few months. Your procedures should outlive all of them.
Exact requirements, levels and file formats are in the full specification.
These books, teaching frameworks and case studies inform YAIBOS. Each profile links to the original work and explains what we take from it.
Author of 'Building a Second Brain' & Founder of Forte Labs
Developed PARA for organizing information by projects, areas, resources and archives, and CODE for turning captured information into useful work.
In YAIBOS: YAIBOS uses PARA to organize business context by active projects and ongoing responsibilities. Open Markdown storage is a separate YAIBOS design choice.
Sources and details
Founder of Linking Your Thinking
Teaches note-making with linked notes and Maps of Content: notes that help gather, develop and navigate related ideas.
In YAIBOS: README maps and links between notes help people and agents find the relevant context without loading the whole second brain.
Sources and details
Author of 'Arrêtez d'oublier ce que vous lisez !' & Co-Founder of Audeon
Shows how an IPCRA second brain in Obsidian can provide context to Claude Code, which reads, organizes and updates the underlying files.
In YAIBOS: Business context and procedures live in files that an agent can read and update. Obsidian is the current viewer and editor; the files remain usable through other tools.
Sources and details
AI Automation Educator & Creator of the AIS-OS Starter Kit
Teaches Context, Connections, Capabilities and Cadence, with a starter kit for building an AI operating system in Claude Code and Codex.
In YAIBOS: Company context, a tools registry, reusable skills and authorized schedules cover similar needs in YAIBOS. The Four Cs framework is credited to Nate Herk.
Sources and details
Authors of the AI Second Brain Case Study
Reports an internal second brain built from PARA folders, context loaded on demand, connected tools and reusable Markdown skills.
In YAIBOS: Load a small amount of root context first, then open project details when needed. Document useful workflows as reusable skills.
Sources and details
AI Researcher & Educator, Former Tesla AI Director & OpenAI Founding Member
Describes an LLM-maintained wiki that turns curated sources into linked Markdown pages, with workflows for ingestion, questions and maintenance.
In YAIBOS: Keep original sources, preserve useful findings in durable notes and periodically check links and stale claims. YAIBOS extends this to business procedures and role-agent context.
Sources and details
Creator of MemoryOS & AI Adoption Educator
Offers MemoryOS: workspace health checks and persistent agent memory stored in an owner-controlled SQL database.
In YAIBOS: Recurring maintenance is part of the Living Second Brain pillar. MemoryOS is a reference for that concern; it is not an integrated YAIBOS dependency or evidence that YAIBOS implements its monitoring features.
Sources and detailsEvaluating active AI agent platforms, second-brain frameworks, and visual orchestrators against the YAIBOS sovereign small-business standard.
These letter grades (A+, B+, C, etc.) do not measure the overall engineering merit, usefulness, or general value of these projects. Ratings compare YAIBOS Base requirements and optional Cockpit and Team capabilities. These existing grades are qualitative comparisons, not conformance certification for draft v1.1. A missing dashboard or team permission service does not make a Base implementation incomplete. Read the three implementation profiles before comparing tools.
Self-hosted, open-source AI workspace packaging multi-model chat, automations, and an embedded Markdown vault visualization into native cross-platform desktop installers.
Screenshots from the project's README (source).
An open file-based multi-agent harness specification where agents, reusable skills, scheduled tasks, and memory live in plain Markdown, featuring a context-aware modular installer and per-agent security contracts.
The leading open-source modular agent harness built on the Cordis plugin engine with swappable memory, tool, and execution plugins.
Self-hosted, open-source platform for orchestrating AI agents through a visual drag-and-drop canvas with native MCP support, prompt vault, and team RBAC.
AI-native product studio and automation agency deploying custom autonomous agents, onboarding pipelines, and bespoke operational workflows for business operators.
Local-first visual desktop interface built directly on top of DeepSeek Harness (DSH), bringing a consumer-friendly UI to a vibrant open harness.
Screenshots from the project's README (source).
Self-hosted AI search and personal assistant that indexes local Markdown notes and documents.
The premier local Markdown editor with community agent plugins (Smart Connections, Banners, Local LLMs).
All-in-one desktop and enterprise app for chatting with documents and local models.
Visual canvas builders for chaining LLMs, tools, and multi-agent workflows into APIs.
Feature-rich self-hosted chat interface for Ollama, OpenAI, and compatible endpoints.
Code-first multi-agent orchestration libraries (CrewAI, AutoGPT, OpenManus) providing programmatic agent harnesses for developers.
Reading open specifications is great: having an operating system running inside your business is transformative. In the AI Bootcamp at Simple AI Studio, you build your sovereign AI Business OS step-by-step alongside other ambitious founders and small business operators.
Organize your company knowledge into open Markdown files that you own forever. Zero cloud database lock-in.
Turn recurring workflows into portable skills. Test them with your AI harness and define when approval is needed.
Use your existing AI plan through its official harness, within its usage limits, or choose your own API key.
Start from the graphical app of a supported harness. A dedicated skill cockpit is an optional extension.
Install your local desktop harness, configure Model Context Protocol (MCP) servers, and connect your AI subscriptions.
Build your company context in open Markdown and define clear operating rules for your AI.
Capture your processes as portable skills with clear inputs, outputs and approval rules.
Run and improve your skills in the chosen harness. Estimate time saved and identify useful next steps.
The spec is a public draft under CC-BY-4.0. It defines conformance levels, the skill file format, the tools registry and the safety rules. Comments and pull requests are welcome.
Get weekly, high-signal breakdowns of practical AI agent setups, Personal Knowledge Management (PKM) methods, team second brains, and open harness specifications. Curated directly by Samuel Michelot. No spam, zero AI fluff.
Sent once a week. Unsubscribe anytime with one click.
Have feedback on the YAIBOS specification, want to propose a new implementation for the scorecard, or exploring a sovereign AI harness for your business?
Leave an asynchronous video, voice note, or text message through MyVideoAsk. No calendar friction, zero meeting fatigue.
Send an asynchronous video message, audio note, or text directly to discuss the YAIBOS specification, propose implementations, or collaborate.