# The AI Agent Factory

> The spec-driven blueprint for building and monetizing AI Workforces: reliable Digital FTEs you can trust, deploy, and scale. A course that teaches domain experts and engineers to manufacture, package, and sell AI Workers. Published 2026 by Panaversity.

The AI Agent Factory teaches how to manufacture, package, and monetize AI Workers: the role-based AI agents that staff the AI-Native Company. It runs from critical-thinking foundations through agent fundamentals, real business domains (finance, banking, legal, sales, operations), spec-driven engineering, and production deployment. The thesis: in the AI era the most valuable companies won't sell software, they'll manufacture AI employees; the SaaS era sold subscriptions, the Agent Factory era sells results. The book is grounded in the February 2026 "SaaSpocalypse," when nearly $1 trillion was erased from software stocks after Anthropic's Claude Cowork showed autonomous agents can do the professional work that once justified $200/month SaaS subscriptions.

**The Crash Courses in Getting Started are the primary surface and the place to start**: self-contained primers that carry the real learning, listed first as standard sections, followed by the reference catalogs that operationalize the thesis.

Key terms used throughout: a **Digital FTE** (also AI Worker, AI Employee) is an AI employee working under human supervision; **Spec-Driven Development** treats specs as the source of truth and has AI generate code from them; an **Agent Skill** is reusable expertise packaged as a `SKILL.md` file; **MCP** (Model Context Protocol) is the universal tool-connectivity standard. The full glossary is linked under Start Here.

## Start Here

The conceptual on-ramp. Read these before the courses; they define every term the rest of the record assumes.

- [Getting Started: Crash Courses (section index)](https://agentfactory.panaversity.org/docs/getting-started): The landing page for the fast 80% path, with the full course sequence, audience paths, and time estimates.
- [How to Learn from This Book](https://agentfactory.panaversity.org/docs/how-to-learn-from-this-book): How to use the introductions, crash courses, references, quizzes, flashcards, social annotations, and AI tutoring: five layers, one loop.
- [Thesis](https://agentfactory.panaversity.org/docs/thesis): The architectural argument that AI-native companies manufacture AI employees rather than sell software, composing tools and spawning agents to deliver outcomes at scale.
- [Thesis (Plain English)](https://agentfactory.panaversity.org/docs/thesis/plain-english): The same thesis in short sentences for readers new to AI or business terminology, and for readers whose first language is not English.
- [The Agent Is the Operating Layer](https://agentfactory.panaversity.org/docs/ai-operating-layer): The interface argument: the agentic era dissolves not just SaaS but the app-on-OS model, replacing it with a layer where humans state intent and agents do the work.
- [What You Carry In](https://agentfactory.panaversity.org/docs/what-you-carry-in): The ownership argument: what a person can actually own when the model is rented, the runtime keeps changing, and the method is given away in a book. Five candidates fail the test.
- [Why AI Is Non-Negotiable](https://agentfactory.panaversity.org/docs/why-ai-is-non-negotiable): The nine core objections to AI answered, from mass unemployment to surveillance, arguing that opting out has never been a viable strategy.
- [Preface: The Right Side of the Line](https://agentfactory.panaversity.org/docs/preface-agent-native): Opens with the 2026 SaaSpocalypse, when roughly a trillion dollars of software value evaporated; this book is the map for those who build and own AI Workers.
- [The Roles This Book Trains](https://agentfactory.panaversity.org/docs/roles-this-book-trains): The job map of the agentic era: Forward Deployed Engineer, Digital FTE Builder, AI-Native Company Architect, and how far this book carries you toward each.
- [Glossary](https://agentfactory.panaversity.org/docs/glossary): Plain-English definitions of the AI, agent, and software terms used throughout the book; start with the Top 30 you must know first.

## Crash Courses (self-contained, start here)

The fast path into the book, and since the Just-in-Time chapter tier was retired these courses are the whole reader-facing surface. Short, high-leverage primers built to make a reader productive in days, not months. Every reader takes the six Foundations courses first, picks a General Agent (Claude Code or OpenCode for engineers, Cowork or OpenWork for knowledge workers), then follows Mode 1 (use general agents to do your work) or Mode 2 (manufacture AI Workers that do the work for you). The fastest route to a shipped Digital FTE is Foundations (1-6), then Agentic Coding (9), Build AI Agents (27), and From Agent to Digital FTE (29), about 15 hours of focused work. Numbering below follows the live sidebar order.

### Section Indexes

Navigational landing pages, one per group below. Each explains the sequence, the audience, and why the courses sit in that order.

- [Foundations: Start in the Browser](https://agentfactory.panaversity.org/docs/foundations): The six skills every reader needs before using or building AI agents, all learnable in a browser.
- [General Agents](https://agentfactory.panaversity.org/docs/general-agents): Pick the general agent you will drive for the rest of the book, then learn to direct it: spec, loop, harness, graph, checker, runtime.
- [Personal Agent Harnesses](https://agentfactory.panaversity.org/docs/personal-agent-harnesses): Open-source AI Employees you run and own yourself, taught through OpenClaw and Hermes. Optional bridge.
- [Mode 1: Problem-Solving](https://agentfactory.panaversity.org/docs/mode-1-problem-solving): Solve a real problem once with a general agent, and do it well: diagnose, solve cleanly, then cross a proven repeat into a permanent worker.
- [Mode 2: Manufacturing](https://agentfactory.panaversity.org/docs/mode-2-manufacturing): Build the worker once and it works without you: gather the building blocks, assemble one trustworthy worker, then scale it into a governed workforce.

### Foundations (Everyone)

- [1. What AI Actually Is](https://agentfactory.panaversity.org/docs/what-ai-actually-is-crash-course): A no-math, no-code mental model of language models: nine ideas explaining why AI predicts rather than looks up, sounds certain when wrong, and miscounts letters. About 45 minutes.
- [2. AI Prompting in 2026](https://agentfactory.panaversity.org/docs/ai-prompting-2026): A 13-concept, ~45-minute primer on using Claude, ChatGPT, and Gemini well in 2026: context, reasoning modes, deep research, multimodal, and AI desktop apps.
- [3. Markdown In, HTML Out](https://agentfactory.panaversity.org/docs/markdown-html-crash-course): A 14-concept primer on the two document languages of agentic work: writing Markdown specs precise enough for a machine, and demanding HTML output rich enough for a human. Prereq: Course 2.
- [4. Code You Never Write](https://agentfactory.panaversity.org/docs/code-you-never-write-crash-course): A 13-concept primer on getting AI to write, run, and verify code you never read, across Claude.ai, Claude Code, OpenCode, Cowork, and OpenWork. Prereq: Courses 2 and 3.
- [5. Skills & Connectors](https://agentfactory.panaversity.org/docs/skills-connectors-crash-course): A no-code primer on the two upgrades that turn a chat box into a coworker: Skills teach AI a task once; Connectors give safe, MCP-scoped access to your apps. About 75 minutes.
- [6. How to Think in the AI Era](https://agentfactory.panaversity.org/docs/how-to-think-ai-era): Six thinking disciplines against six AI failure modes, with one rule: the deliverable is documented evidence of thinking. About a working day of exercises.

### General Agents (Pick Your Co-Worker, Then Learn to Direct It)

- [7. General Agents on the Web](https://agentfactory.panaversity.org/docs/general-agents-web-crash-course): The front door to the section: direct your first general agent from a browser tab with nothing installed. Remote sessions, the account spine, files in three tiers, and the approval that reaches your phone. Everyone starts here.
- [8. Open Source LLMs: Your Laptop, Your Server, the Cloud](https://agentfactory.panaversity.org/docs/using-open-source-llms): Run open models at all three scales: Ollama on your own laptop, vLLM on a GPU box serving many people, and OpenRouter for the frontier open models nobody can self-host. Your agent is a harness plus a swappable brain.
- [9. Agentic Coding Crash Course: Claude Code and OpenCode](https://agentfactory.panaversity.org/docs/agentic-coding-crash-course): A 90-minute, 15-concept tour of Claude Code and OpenCode: same vocabulary, slightly different keybindings, skills transfer cleanly. The general-agent starting point for engineers, and the book's context-engineering course.
- [10. Cowork Crash Course](https://agentfactory.panaversity.org/docs/cowork-crash-course): A 90-minute, 15-concept primer on Claude Cowork and OpenWork: delegating real desktop knowledge work, the autonomy ladder, prompt-injection defenses, and the plan-review habit. The starting point for knowledge workers.
- [11. Website Design: Made, Not Generated](https://agentfactory.panaversity.org/docs/website-design-crash-course): A 15-concept applied build course that points the tools at one real job: design, verify, and ship a client website, single-page and multi-page, that looks made rather than generated. Taste as a file the model loads, a four-lane media pipeline, screenshots used as tests, and a live URL. Prereq: Course 9.
- [12. Spec-Driven Development](https://agentfactory.panaversity.org/docs/spec-driven-development-crash-course): A 13-concept primer on agreeing what to build before generating how: the project constitution and four phases, run in claude.ai, Claude Code, and OpenCode. About 90 minutes.
- [13. Loop Engineering](https://agentfactory.panaversity.org/docs/loop-engineering-crash-course): A 15-concept primer on designing the loop that prompts your agent for you: heartbeat, worktree isolation, skill, maker-checker split, connector, and state spine. About 2 hours; prereq Course 9.
- [14. Harness Engineering](https://agentfactory.panaversity.org/docs/harness-engineering-crash-course): A 12-concept primer on the layer that turns a model into an agent you can trust: the five verbs (constrain, inform, verify, correct, escalate), permission rules, sandboxes, hooks, typed output, the four failure classes, and the ratchet. About 2 hours; prereq Courses 9 and 13.
- [15. Graph Engineering](https://agentfactory.panaversity.org/docs/graph-engineering-crash-course): The memory half of autonomy: stop leaving findings in transcripts and write them as typed, connected records any later agent can query. Two graphs never collapsed into one, a receipt on every claim, and an honest account of when not to build a graph at all. Prereq: Course 13.
- [16. Trusting the Checker: An Evals Crash Course](https://agentfactory.panaversity.org/docs/trusting-the-checker-crash-course): A 12-concept primer on evals at operating size: grade the pass rate not the run, the three depths, the golden set from real failures, calibrating the judge against your own blind grading, the regression gate, drift, and Goodhart's law. Folder of cases, shell, and jq, no frameworks. Prereq: Courses 9 and 14.
- [17. Leaving the Laptop: A Runtime Crash Course](https://agentfactory.panaversity.org/docs/leaving-the-laptop-crash-course): The runtime decision that closes the section: where a proven loop should live once it has earned trust (a cloud schedule, a managed runtime, or your own process) without losing the track record it earned.

### Personal Agent Harnesses (Optional Bridge)

- [18. OpenClaw with General Agents](https://agentfactory.panaversity.org/docs/openclaw-with-general-agents): A 90-minute, 6-scenario hands-on course where your general agent installs and configures a Personal AI Employee on OpenClaw: custom skill, MCP tool, heartbeat, ACP-spawn. Prereq: Course 9.
- [19. Hermes with General Agents](https://agentfactory.panaversity.org/docs/hermes-with-general-agents): A 90-minute, 6-scenario course where your general agent installs and runs a self-improving AI Employee on Hermes, one that writes its own skills. Prereq: Course 9.

### Mode 1: The Problem-Solving Track

- [20. Is This an Agent Problem?](https://agentfactory.panaversity.org/docs/is-this-an-agent-problem): Three gates to run in ten minutes before you open an agent: is this even an agent job, will you do it once or every week, and do you know what finished looks like. The Mode 1 triage.
- [21. Problem Solving with General Agents](https://agentfactory.panaversity.org/docs/problem-solving-crash-course): A 90-minute, 7-principle crash course in the operating discipline that turns any general agent (Claude Code, OpenCode, Cowork, OpenWork) into a tool you can ship real work on. The Mode 1 governing course.
- [22. From One-Off to Worker: The Handoff to Manufacturing](https://agentfactory.panaversity.org/docs/from-one-off-to-worker): When a task you keep solving by hand is ready to become a permanent worker: the one signal that it is time to cross, and the four parts of your Mode 1 solution that carry over into Mode 2.

### Mode 2: Manufacturing, Phase 1 (Building Blocks)

- [23. Python in the AI Era](https://agentfactory.panaversity.org/docs/python-crash-course): A read-before-you-write primer for non-coders: read, predict, test, and verify the Python your agent generates via PRIMM-AI+ and the TDG loop. 17 concepts; prereq Course 9.
- [24. Connector-Native Apps](https://agentfactory.panaversity.org/docs/connector-native-apps): Build and ship a remote MCP server whose customer is an AI: a free-tier Claude user adds it with one pasted URL and one Authorize click. The pre-loop build course.
- [25. Plugins for AI Agents](https://agentfactory.panaversity.org/docs/plugins-crash-course): Build and ship one real plugin that extends a coding agent you do not own (Claude Code or OpenCode). The bundle is host-specific; the portable skill inside it is not.
- [26. AI Identity: Human Sign-In and Agent Access](https://agentfactory.panaversity.org/docs/ai-identity-crash-course): The identity and access layer of the Manufacturing path: own your sign-in (email and social login, sessions, two-factor, OAuth/OIDC), then decide what an agent is allowed to reach.
- [27. Build AI Agents Crash Course](https://agentfactory.panaversity.org/docs/build-agents-crash-course): A 16-concept primer on the OpenAI Agents SDK: agent loop, tools, sessions, streaming, handoffs, guardrails, tracing, human approval, and sandboxed deployment. 90 minutes to read, 4-6 hours to build. Prereq: Course 9.

### Mode 2: Manufacturing, Phase 2 (Build Workers)

- [28. Give Your AI Searchable Context: RAG on Postgres with pgvector](https://agentfactory.panaversity.org/docs/postgres-ai-crash-course): A 15-concept primer on giving your AI searchable context: direct your agent to build RAG on Neon plus pgvector with schema, embeddings, hybrid search, and a read-only MCP server. Prereq: Course 9.
- [29. From Agent to Digital FTE](https://agentfactory.panaversity.org/docs/digital-fte-crash-course): A 4-hour workshop turning a basic agent into a durable Worker: portable Skills, Neon Postgres plus pgvector as system of record, MCP between them. 15 concepts, 8 decisions; prereq Course 27.
- [30. From Digital FTE to Production Worker with a Nervous System](https://agentfactory.panaversity.org/docs/ai-agent-nervous-system-crash-course): A 90-minute, 15-concept course wrapping your Digital FTE in an Inngest envelope: durable execution, event triggers, step memoization, concurrency, replay, and human-in-the-loop gates. Prereq: Course 29.

### Mode 2: Manufacturing, Phase 3 (Scale the Workforce)

- [31. Human-Agent Teams](https://agentfactory.panaversity.org/docs/human-agent-teams-crash-course): The operating model for running humans and Digital FTEs as one team: work in the open, one roster with clear roles, a north star, and trust that grows with evidence.
- [32. Designing Agent Experiences](https://agentfactory.panaversity.org/docs/designing-agent-experiences-crash-course): Designing agentic products, with a hands-on MCP Apps tutorial. An agent product has two users at once: a human who must trust it and an agent that must operate it.
- [33. Building a Workforce with Paperclip](https://agentfactory.panaversity.org/docs/workforce-with-paperclip-crash-course): A 90-minute hands-on course where your agent stands up Paperclip, the MIT-licensed AI-native company control plane: hire a keyless Worker, file board approvals, meter a budget, audit via SQL. Prereq: Course 29 or 18.
- [34. From Fixed to Dynamic Workforce](https://agentfactory.panaversity.org/docs/dynamic-workforce-crash-course): A half-day, 15-concept and 7-decision workshop where a Paperclip workforce detects a capability gap, drafts a hire, and provisions a new Worker under policy. Closes Invariant 6; prereq Course 33.
- [35. From Founder Bottleneck to Owner Delegate](https://agentfactory.panaversity.org/docs/identic-ai-crash-course): A 4-hour, 15-concept and 7-decision workshop building an Owner Identic AI on OpenClaw that clears routine approvals inside a signed delegated envelope. Closes Invariant 2; prereq Course 34.
- [36. Eval-Driven Development for AI Employees](https://agentfactory.panaversity.org/docs/eval-driven-development-crash-course): The discipline that closes the manufacturing arc: the nine-layer evaluation pyramid and a four-tool stack (OpenAI Agent Evals, DeepEval, Ragas, Phoenix). Four tracks, 15 concepts.
- [37. Deploy Your Agent Harness to the Cloud](https://agentfactory.panaversity.org/docs/deploying-agents-crash-course): Ships the manufacturing track: the harness/sandbox split on FastAPI and Azure Container Apps, Neon, Cloudflare R2, a code sandbox, observability, and the eval suite as a CI gate. Four tracks, 17 concepts; prereq Course 36.
- [38. Choosing Agentic Architectures](https://agentfactory.panaversity.org/docs/choosing-agentic-architectures-crash-course): A conceptual crash course on pattern selection: five questions map your task to one of four core patterns plus reflection. Choose by architectural fit, not by what looks impressive. Four tracks.
- [39. Payment-Enabled Agents: ACP, AP2, x402, and MPP](https://agentfactory.panaversity.org/docs/payment-enabled-agents-crash-course): The four protocols that let agents move money (ACP, AP2, x402, MPP) as composable layers, not rivals, built in OpenAI Agents SDK code. 19 concepts, four tracks; prereq Course 27.

### Companions & References (within Getting Started)

- [40. Which AI Employees Should You Use in 2026?](https://agentfactory.panaversity.org/docs/which-agents-2026): Five AI tools matched to who you are and what you need; find your starting point in under a minute.
- [41. Cheatsheets](https://agentfactory.panaversity.org/docs/cheatsheets): Interactive quick-reference cards for the book's key tools: Claude Code, the Claude collaborative workspace (Cowork), and OpenClaw.
- [42. Agentic Engineering Fundamentals](https://agentfactory.panaversity.org/docs/agentic-engineering-crash-course): A 45-minute primer on the engineering discipline beneath the section: smart zone versus dumb zone, jagged intelligence, the six failure modes of AI coding, and the seven-stage grill. Optional companion past Course 29.

## Reference Catalogs & Companions

Reference material that operationalizes the thesis into catalogs you can sell from, plus the ecosystem's own architecture and business model, the playbooks for building a vertical on it, and the certification program.

- [AI Workers](https://agentfactory.panaversity.org/docs/ai-workers): The hub for four catalogs that operationalize the thesis: which AI Workers exist, and how to sell, price, and market them.
- [The AI Worker Catalog](https://agentfactory.panaversity.org/docs/worker-catalog): A living reference of role-based AI Workers (Digital FTEs) across business verticals: what they do, read, write, and who supervises them.
- [The AI-Native Finance Catalog](https://agentfactory.panaversity.org/docs/finance-catalog): How AI-native companies price, forecast, and model unit economics when costs scale with usage, from a year-one prescription to advanced financial architecture.
- [The Sales Catalog](https://agentfactory.panaversity.org/docs/sales-catalog): Motions for selling AI Workers, sequenced from buyer-led self-serve through vendor-led, outcome-led, and partner-led enterprise deals.
- [The Marketing Catalog](https://agentfactory.panaversity.org/docs/marketing-catalog): Motions for building demand for AI-native products: awareness, positioning, and pipeline, starting with founder-led content before any paid marketing.
- [AI-Native Companies](https://agentfactory.panaversity.org/docs/ai-native-companies): The company-level view of the Agent Factory: how an existing business transforms into an AI-native company and how a founder builds one from nothing.
- [The AI-Native Transformation: From Adoption to Advantage](https://agentfactory.panaversity.org/docs/ai-native-transformation): Transformation postures for an existing organization moving from AI adoption to durable advantage.
- [Founding an AI-Native Startup](https://agentfactory.panaversity.org/docs/ai-native-companies/how-to-found-an-ai-native-startup): The startup-founding playbook for building an AI-native company from zero.
- [How to Get Paid in the Agentic AI Era](https://agentfactory.panaversity.org/docs/how-to-get-paid-agentic-ai-era): The monetization capstone: how the book's skills turn into money in eight stations, and the three ways they earn, namely a job, freelance retainers, or a domain startup.
- [Courses & Certifications](https://agentfactory.panaversity.org/docs/certifications): A five-course certification program from general-agent problem-solving to shipping AI-Native Companies, earning the Certified Agentic AI Architect credential.

### The Agent Factory Ecosystem

The ecosystem's own architecture: the records it runs on, the products built on them, and the playbooks for building a vertical of your own.

- [The Agent Factory Ecosystem](https://agentfactory.panaversity.org/docs/ecosystem): One System of Record, and the products that teach and build on it, reaching hundreds of thousands of learners and developers at almost no AI cost.
- [Agent Factory System of Record](https://agentfactory.panaversity.org/docs/ecosystem/system-of-record): The book's content as a system of record, served over MCP. Connect any AI agent or worker (Claude, ChatGPT, Claude Code, Cowork, or your own) and ground it in the book.
- [Zia Tutor AI](https://agentfactory.panaversity.org/docs/ecosystem/zia-tutor-ai): A personal learning agent, built as a digital twin of Zia Khan, that teaches the Agent Factory, remembers your progress, and continues where you stopped. It lives inside the AI agent you already use, including Claude.
- [Zia Developer AI](https://agentfactory.panaversity.org/docs/ecosystem/zia-developer-ai): Your AI build partner inside Claude Code and OpenCode: describe the problem, and it picks the architecture from the book, writes the spec, and builds the agent.
- [The FDE AF Model](https://agentfactory.panaversity.org/docs/ecosystem/fde-af-model): The five-layer platform and business model the ecosystem runs on: what each layer is, the one law that connects them (repeated work moves down), and where a graduate earns at every layer above the foundation.
- [Choosing Your Vertical](https://agentfactory.panaversity.org/docs/ecosystem/choosing-your-vertical): The selection discipline for a domain startup at Layer 3: a six-question screen for selling ease, eight tests for whether the domain can carry a trio, the beachhead rule, and four launch gates, with two honest exits for candidates that fail.
- [Designing the Vertical System of Record](https://agentfactory.panaversity.org/docs/ecosystem/designing-the-vertical-sor): What goes inside the first thing you build, derived rather than copied: workflow archaeology, the three-bin sort (keep what law and trust require, redesign what human limits required, delete what old technology required), the source hierarchy, the two content classes, and the thin slice that proves it.
- [The System of Context](https://agentfactory.panaversity.org/docs/ecosystem/system-of-context): The federated layer above a portfolio of governed records. A System of Record decides what is true; a System of Context decides what arrives. Authority routing, the context packet, what may be indexed and what must be asked, permission before the model, provenance, and conflict as a result rather than a failure.

## Audience

- **Subject matter experts** (CFOs, CAs/CPAs, lawyers, compliance officers, supply chain managers, HR directors): translate decades of professional judgment into SKILL.md libraries and supervise AI rather than fear replacement.
- **Engineers and platform builders**: build, operate, and scale AI employees with the same discipline applied to other technical systems.
- **Enterprise leaders**: understand the maturity model, the build-vs-buy framework, and the deployment patterns that turn the Agent Factory thesis into operating reality.

## Key Terminology

- **The Agent Factory**: the spec-driven, human-supervised, Claude-Code-powered *process* by which AI Workers are designed, manufactured, and deployed. A practice you adopt, not a product you buy.
- **The AI-Native Company / Agentic Enterprise**: the *output* of the Agent Factory: a firm staffed by AI Workers, coordinated by a management plane, directed by humans at the edge.
- **AI Workers / Digital FTEs / Digital Workers**: the role-based agents that staff the AI-Native Company. Hired, assigned, rostered, retired.
- **Manufacture / Package / Monetize**: the triad that defines the Agent Factory operator's loop: manufacture AI Workers from specs, package them into deployable products, monetize results rather than subscriptions.

## Optional

Low-priority meta pages. Following the llms.txt convention, "Optional" means an agent on a tight context budget can skip this whole section; it is not a judgment of importance.

### Meta

- [About](https://agentfactory.panaversity.org/docs/about): The Agent Factory ecosystem, delivered as a book, an AI tutor, an AI building partner, and derivative books.
- [What's New](https://agentfactory.panaversity.org/docs/whats-new): Meaningful updates to the book, grouped by day and populated automatically from merged pull requests.