# OpenCortyx: Universal Knowledge Infrastructure

> OpenCortyx is an open knowledge layer where anything can become connected knowledge, and anyone or any AI agent can retrieve the context they need from it.

OpenCortyx brings together data, documents, code, conversations, and agent memories into a single connected knowledge layer. When you or an AI agent asks a question, OpenCortyx traces connected relationships to retrieve exact grounded context without noise or guesswork.

## Core Resources

- [OpenCortyx Homepage](https://www.opencortyx.com/): Official platform overview and architecture highlights.
- [Technical Documentation](https://www.opencortyx.com/docs/): Complete guide to the graph model, PostgreSQL 16 substrate, pgvector search, and MCP integration.
- [Architecture Flow](https://www.opencortyx.com/#how-it-works): Visual walkthrough of data ingestion, relationship mapping, and context retrieval.
- [Early Access Application](https://www.opencortyx.com/#cta): Request developer beta onboarding.

## Architecture and Design

OpenCortyx connects scattered information into an open substrate:

```
                      +-------------------+
                      |   Your World      | (Databases, Docs, Code, Memory, APIs, Events)
                      +---------+---------+
                                | Ingestion & Relations
                                v
                      +-------------------+
                      | OpenCortyx Engine | (Knowledge, Relations, Context)
                      | - Entities & Nodes| (Typed relational graph)
                      | - Documents & Data| (PostgreSQL + JSONB + Vectors)
                      | - Rules & Context | (Grounding and provenance)
                      +---------+---------+
                                | Targeted Retrieval
                                v
                      +-------------------+
                      | Context Engine    | (Finds relevant connections & paths)
                      +---------+---------+
                                | Model Context Protocol
                                v
                      +-------------------+
                      | Humans & Agents   | (Engineers, Claude, Cursor, Custom Agents)
                      +-------------------+
```

### Core Principles

1. Knowledge has no fixed shape: Information can be a database row, document, memory, event, claim, or observation. OpenCortyx lets you define what matters and connect it.
2. Memory is one kind of knowledge: Rather than treating agent memory as an isolated chat buffer, OpenCortyx connects agent memories, human notes, and system databases into one unified model.
3. Context on demand: When you or an agent asks a question, OpenCortyx traces relationships to retrieve only relevant context, saving tokens and eliminating hallucination.
4. Built for people and tools: OpenCortyx gives your team and all your tools a shared knowledge layer rather than having each tool run in a disconnected silo.
5. Standard interfaces: Native Model Context Protocol (MCP) support for IDEs, coding bots, and autonomous agents.
6. Local privacy: Self-hosted and run locally with PostgreSQL. Your data stays entirely in your environment.

## Model Context Protocol (MCP) Integration

Configure your IDE client (Cursor, Claude Desktop, Windsurf, Zed) to connect to OpenCortyx:

```json
{
  "mcpServers": {
    "opencortyx": {
      "type": "sse",
      "url": "http://localhost:8001/sse",
      "headers": {
        "X-API-Key": "oc_live_key_production"
      }
    }
  }
}
```

### Standard MCP Tools Exposed

- read_context: Query the knowledge engine for grounded facts, relations, and source citations.
- write_relationship: Insert or update entities and typed connections in the knowledge graph.
- update_memory: Record agent decisions, observations, and persistent context nodes.
- verify_provenance: Check source references and provenance for retrieved context.

## External References

- [Technical Reference Manual](https://www.opencortyx.com/docs/): Exhaustive technical reference and schema documentation.
- [Canonical Site Root](https://www.opencortyx.com/): Official production deployment.