Nathan Lambert hosts Nufar Gaspar to introduce Agent OS, AIDB's latest free training program for building agentic operating systems. Unlike ClawCamp which focused exclusively on OpenClaw, Agent OS is designed to be platform and model neutral, working across Claude Code, Cursor, Codex, and other agentic tools.

The conversation explores how agentic tools are converging on similar capabilities, making the underlying system architecture more critical than specific tool selection. Nufar presents a seven-layer framework for building extensible agentic systems focused on knowledge work rather than coding, using a chief of staff agent as the primary example throughout the discussion.

The Convergence of Agentic Tools and Platform Independence

Every agentic tool is becoming every other tool - Cursor added agents, Claude Code added memory systems, OpenClaw reads files, and Codex runs in the background, all converging on the same capabilities.

Tool choice matters less than the underlying system that captures how you work, what you know, and what you need AI to do for you.

All agentic tools read the same human-readable text files that define identity, knowledge, capabilities, memory, and connections - making your work portable across platforms.

The Seven-Layer Agent OS Architecture

Layer 1 - Identity: Defines who you are, communication style, values, and rules enforced every time the agent interacts (called Sol in OpenClaw, agents.md in Cursor, claude in Claude Code).

Layer 2 - Context: Contains 3-5 focused, single-page files covering your team, product, customers, quarter, and stakeholders - updated when things change rather than massive static documents.

Layer 3 - Skills: Reusable instruction sets for repeated workflows like 'when I say X, do Y process using Z sources and produce output in this format.'

Layer 4 - Memory: Understanding your tool's memory limitations and deliberately capturing major decisions, priority changes, and relationship context beyond automatic memory.

Layer 5 - Connections: Read-only access to real systems (email, calendar, Slack) using MCPs, CLI tools, or APIs - write access only after weeks of trusted behavior.

Layer 6 - Verification: Quick 3-5 checks for tone, facts, and accuracy plus periodic system audits to identify unused skills or stale context files.

Layer 7 - Automations: Scheduled tasks that run without supervision, starting with draft outputs for review rather than direct external communication.

Building Your Chief of Staff Agent

A chief of staff agent reviews inbox, prepares meeting pre-reads, tracks commitments, flags blind spots, drafts weekly updates, and eventually manages other agents.

Essential context files include stakeholders (who reports to you, key partners, what each cares about), strategy and priorities, and operating principles for decision-making.

Key skills include pre-read generation, daily briefings, voice matching, and commitment tracking across calls and communications.

"Context creation is the single fastest path to AI value" - when people understand this, they stop asking what tool to use and start asking what knowledge isn't written down - Nufar.

Security and Risk Management for Agentic Systems

Start with read-only access to external systems and add write permissions only after observing agent behavior for several weeks.

"The worst thing that happens isn't that it fails, it's that it works confidently and wrongly, and you ship the output before you notice" - Nufar.

Real incidents already occurring include agents with Slack access sharing private notes and opinions about colleagues when team members chat with them.

Use least privilege connections, coordinate with IT teams for work systems, and avoid creating cautionary tales for your organization.

The Compounding Returns of Agent OS Investment

First agent takes a weekend to build while constructing the OS foundation, but second agent takes only an afternoon by inheriting existing layers.

Nufar's personal system includes Chloe (chief of staff) plus specialist agents for content, technical building, and platform work, all sharing the same OS foundation.

The Agent OS framework represents evolution from disconnected AI projects to integrated systems that persist across tool changes and model updates.

"People who build that foundation now will have it compound from here on after, and everyone else will keep starting over with new tools" - Nufar.

Resources Mentioned

The Toyota Way, Second Edition 14 Management Principles from the World's Greatest Manufacturer

primarily on knowledge work. So whether it's strategy, communication, operations, decision making, research management, and anything that the knowledge worker does, that is where the most professiona

Station Agent Exam Prep 2026-2027 Complete Study Guide with Practice Tests, Proven Strategies, and Real-World Scenarios for Guaranteed Success

ff maybe took you a weekend, but the second agent that is built on top of this system, maybe it's a research agent or a board prep agent, that takes you an afternoon because it inherits everything tha