Glossary
What is Context OS?
A Context OS is a structured, queryable layer of company knowledge that AI tools use to answer questions, ground decisions, and act with shared context. It sits between scattered documents and the AI tools that need to reason over them, turning raw information into source-grounded, role-aware, queryable infrastructure.
Last updated:
Origins
Where the term comes from
The term emerged from a few overlapping conversations in 2024 and 2025: Andrej Karpathy's 'LLM Wiki' methodology for treating personal and team knowledge as AI-readable wikis; Anthropic and LangChain's framing of 'context engineering' as the practice of giving AI agents the right context at the right time; and the developer ecosystem around agent memory, retrieval, and structured knowledge for LLMs. A Context OS extends those ideas into infrastructure: where a wiki is for one person and context engineering is a practice, a Context OS is the system that operationalizes both for companies, with shared access, source-grounded answers, and AI-tool integration that survives team turnover and AI vendor changes.
Capabilities
What Context OS does
Canonical knowledge storage
Stores company knowledge in a single source of truth: typically markdown files on disk, not behind a cloud API.
Typed graph structure
Adds types to knowledge: concepts, patterns, people, decisions, customers, with status, confidence, and relationships between them.
Programmatic query interface
Exposes the structured knowledge to AI tools through MCP, API, or other interfaces so AI agents can traverse the graph and answer with grounded context.
Role-aware access
Controls what each teammate or AI agent can see, edit, or audit, typically through viewer, maintainer, and admin roles.
Source attribution
Every claim carries source, status, and ownership so the team can tell what is proven, current, and safe to reuse versus what is still emergent.
Compounding insight
New information links to existing concepts, strengthening the network rather than getting lost in a separate document.
Distinctions
Context OS vs adjacent concepts
Context OS is often confused with related but distinct ideas. Here is how it differs.
| Concept | What it is | How Context OS differs |
|---|---|---|
| Knowledge graph | A data structure. Typed nodes and relationships. | The system that surrounds and operates the graph. A Context OS uses a knowledge graph as one component, plus ingestion, query, access, and AI integration. |
| Wiki / company docs | Pages for humans to read. No types, no programmatic query, no AI integration. | Structured for both humans and AI. Queryable in plain English by AI agents. Carries types, source, confidence, and access roles. |
| AI memory | What one AI assistant remembers across a single conversation or user. | What the company remembers across people, time, AI tools, and team turnover. Persistent, shared, role-controlled. |
| RAG | A retrieval pattern that fetches relevant text chunks at query time. | The source of what gets retrieved. A Context OS is the structured knowledge layer; RAG is one way to query it. |
| Notion / Confluence | Document workspaces. Files behind a cloud API. | Markdown canonical on disk. The graph and MCP server sit on top. No cloud-API tax, no vendor lock-in. |
Who uses it
Who uses Context OS
Founder-operators whose AI knowledge is locked inside one vendor's chat surface and cannot be searched as a graph, traversed across decisions, or used outside that AI's chat. Teams whose company knowledge is scattered across docs, calls, and individual heads. Companies where non-technical members need to query company context across multiple AI tools without touching git.
FAQ
Common questions about Context OS
Is a Context OS the same as a knowledge graph?
No. A knowledge graph is a data structure: typed nodes connected by typed relationships. A Context OS is the system around the graph: ingestion, query interface, role-based access, source attribution, and AI tool integration. The graph is one component of a Context OS.
Is a Context OS the same as a wiki?
No. A wiki is a collection of pages for humans to read. A Context OS is a structured layer that both humans and AI tools can query. A wiki has no types, no programmatic query interface, and no role-aware access for AI agents.
How is a Context OS different from AI memory?
AI memory is what one AI assistant remembers across a conversation. A Context OS is what the company remembers across people, time, and AI tools. AI memory dies when the conversation ends; a Context OS persists, compounds, and is shared.
How is a Context OS different from RAG?
RAG (retrieval-augmented generation) is a pattern for fetching relevant text chunks at query time. A Context OS is the source of what gets retrieved: the structured, source-grounded knowledge layer that a RAG system can query against.
Who needs a Context OS?
Founder-operators whose AI projects help one person but never become shared infrastructure. Teams whose company knowledge is scattered across docs, calls, and individual heads. Companies where non-technical members need to query company context without touching git.
Related terms
Keep going
See Context OS in practice
DearTech-OS is a Context OS for founder-operators. Explore the product or talk through whether one is right for your team.