How to describe harperOS.
How we describe harperOS, for partners, press, analysts and the team.
The operating system for companies that run on AI.
harperOS holds a company’s knowledge and runs the agents that act on it.
It sits above the AI tools a company already uses and organizes them around the work, not the other way around.
Built on Claude Code as the core infrastructure. Tool- and model-agnostic by design.
Three connected layers.
Every harperOS deployment organizes around the same three layers.
Each layer compounds the others.
Knowledge
Agents
Workflows
Every agent is specialized.
harperOS agents are not generic copilots.
Each one is built around five components that make it an expert in a specific function — and keep it sharpening over time.
Dedicated knowledge
Dedicated skills
Tool access
Enforcement & rules
Feedback loop
From decision to delivery.
Every harperOS deployment follows the same three-step path.
Decide where to start
Pick the first domain — the function that creates the highest leverage. Usually a department feeling the most pressure or sitting on the most underused knowledge.
Live OS week one
Within seven days, the operating system is live for the chosen domain — knowledge installed, agents deployed, workflows running.
Agents delivered
Production agents work alongside the team from day one. Output per team member doubles. The system compounds as the next domain comes online.
Claude Code as the core infrastructure.
harperOS runs on Claude Code, the foundation for AI agents.
Claude Code provides the runtime, tool surface, and discipline for every agent in the system.
The product is tool- and model-agnostic by design. Claude is the default model; the architecture supports the model and tool choices a company already runs.
Key terminology.
Words to use, and what they mean inside the system.
| Operating system | The full stack that holds knowledge, runs agents, and orchestrates workflows. Not a chatbot. Not a tool. A system. |
| Knowledge layer | The structured, curated, continuously updated record of how the company works — methodology, voice, decisions, examples. |
| Agent | A specialist that owns a defined function. Reads the knowledge layer, executes skills, and uses scoped tools. |
| Skill | A repeatable workflow an agent runs. Versioned, testable, and improvable. The unit of capability. |
| Plugin | A bundle of agents, skills, and knowledge that serves a domain — sales, marketing, finance, account management, etc. |
| Workflow | The orchestration of agents and skills into an end-to-end business outcome. |
| Domain | A functional area of the company — the unit at which harperOS is deployed. |
| Compound intelligence | The accumulation of decisions, patterns, and learnings the system captures and reuses. The reason output keeps getting better. |
Anti-positioning.
Easier to understand harperOS by what it is not.
Not a chatbot
harperOS does not sit behind a prompt box waiting for questions. Agents are scheduled, triggered, and workflow-bound.
Not a model wrapper
Models are interchangeable. The operating system is what compounds — knowledge, skills, rules, and the workflows that connect them.
Not a vertical SaaS
harperOS is horizontal — it runs every department inside one operating layer, not a single function inside one application.
Not a pilot
Built inside live operations from day one. Deployed against real work, with real teams, in production. No sandbox phase.
How to describe it.
Phrases approved for partner, press, and analyst use.
Product name
Category
One-liner
Tagline