01Retrieve relevant text
Retrieval finds content, but structure, constraints, and version context may remain implicit.
PROGRAMMABLE KNOWLEDGE GRAPH INFRASTRUCTURE FOR AI
Beyond retrieval.Make knowledge operable.
Define your own knowledge models. Connect data through relationships. Give AI a structured system it can query, update, and evolve—not just retrieve relevant text from.
- Graph-first
- Custom Models
- MCP + Skills
- Versioned Knowledge
PRODUCT EXPERIENCE / 01
Which components does this decision affect? What supports it?
GraphOntologyQuery
Selected / Decision
Versioned storageAffected modules2
Evidence links1
Current branchmain
Illustrative data and UI concept. Not a screenshot from a running KG OS instance.FIG 01 / KNOWLEDGE GRAPH
01 / WHY KG OS
Search finds information. Knowledge is more than content.
AI needs to understand not only what something says, but how it relates to other things, which state it belongs to, and how it can change.
02Operate structured knowledge
Explicit models and relationships let AI query dependencies, inspect evidence, and make authorized changes.
02 / HOW IT WORKS
From information to operable knowledge.
A software project is one example. You define the model. Operations work against graph data and explicit version states.
01 / DEFINE
Define models
Define nodes, relationships, and constraints for your domain instead of adapting to a fixed schema.
02 / CONNECT
Connect knowledge
Store real entities and relationships as queryable graph data.
03 / OPERATE
Let AI operate
Agents use MCP to query data and submit changes with authorization.
04 / EVOLVE
Keep the history
Manage knowledge changes with commits, branches, diffs, and merges.
AI-NATIVE INTERFACE
Keep your existing agents.
KG OS has no built-in agent. Your AI tools can use MCP to inspect models, query data, and perform authorized operations. Official Skills guide modeling, retrieval, and maintenance workflows.
- MCP
- Skills
- Cypher
- Web
AI / QUERY EXAMPLEREAD-ONLY
// KG OS MCP
kgos_query({
code: "return await evolution.overview();"
})
// Query the real, pinned state
// through the KG OS SDK bridge.03 / CAPABILITIES
Models, data, and history. In one system.
KG OS provides foundational capabilities without prescribing what your domain means. Build the knowledge system your application needs.
01 / ONTOLOGY
Define your own structures
Define entities, relationships, properties, and constraints for your work, with indexes for the information you need to find.
02 / GRAPH & SEARCH
Query knowledge from every angle
Combine graph traversal, structured queries, full-text search, and semantic retrieval.
03 / EVOLUTION
Track how knowledge evolves
Keep models and data in the same history, compare versions, and manage changes with branches and merges.
04 / AI & WEB
AI operates. You stay in control.
Connect your Agent through MCP. Use the Web interface to set up your instance and approve connections.
04 / ARCHITECTURE
Clear responsibilities at every layer.
KG OS focuses on knowledge infrastructure, while agents and the independent graph database engine own their separate responsibilities.
INTELLIGENCEAgent / Skill
Reasoning and decisions
KNOWLEDGEKG OS
Model · MCP · Web
GRAPH ENGINELithograph
Graph · Search · Versioning
STORAGESQLite
Embedded storage
Keep the Agent you already use. One KG OS instance provides MCP and Web access, with a graph database to store and query knowledge.
05 / INSTALL WITH AI
Let your AI Agent handle setup.
Copy the prompt below into an AI Agent with access to your installation environment. It will read the official guide, check your setup, perform permitted actions, and verify the result.
INSTALL.md / OFFICIAL
New deployments need Linux amd64 with Docker; existing instances can be connected directly. This page does not install software. Your Agent checks releases, environment access, and permissions, then explains any steps that need your action.