Name: Cortrix
Author: Cortrix

Defining a New Category

AI agents need more than a vector database. They need a storage engine that speaks their language — natively.

Agent-Native

Not retrofitted for agents — built from day one as the data layer agents think through. Semantic processing, memory, and audit are first-class primitives, not plugins.

Semantic Storage

Beyond vector search. Documents are parsed, chunked, embedded, and indexed in a unified semantic layer — queryable by meaning, not just keywords.

Born to Scale: PB to EB

Architectured for petabyte-to-exabyte workloads from the ground up. Horizontal scaling, namespace isolation, and storage tiering are native — not afterthoughts.

One Engine. Complete Agent Data Stack.

Cortrix consolidates semantic storage, interaction memory, and causal-chain audit into a single C++ engine with REST, MCP, and SQL interfaces.

Semantic Processing Chain (SPC)

Ingest any document — PDF, Word, Markdown — through an automated pipeline: parse, chunk, embed, and index in one step.

Core

Hybrid Query Engine

Combine vector similarity (P-HNSW) with BM25 keyword search via Reciprocal Rank Fusion for best-in-class retrieval accuracy.

Core

AI Interaction Memory

Persistent, searchable conversation memory with session management, semantic recall, context injection, and full audit trail.

Core

Causal-Chain Audit Trail

Full AI decision traceability — not just call indexes, but the actual content at every step. Trace any AI employee's reasoning from input to output for compliance and debugging.

Unique

pgCortrix Extension

Semantic storage as a PostgreSQL extension. Bring agent-native capabilities directly into your existing Postgres infrastructure — no separate service needed.

Integration

Built-in Embedding (bge-m3)

ONNX Runtime integration with bge-m3 model for multilingual embeddings. No external embedding service needed.

Built-in

Connector Ecosystem

Plug into LangChain, Dify, RAGFlow, and any MCP/CLI/REST workflow. Connectors for HTTP upload, directory watch, CDC, and custom data sources.

Ecosystem

MCP Server

Model Context Protocol server exposes Cortrix capabilities to Claude Code, Cursor, and any MCP-compatible AI tool out of the box.

Integration

Multi-Namespace Isolation

Logically separate data by project, team, or tenant. Each namespace has independent storage, indexes, and access control.

Enterprise

REST API + Web UI

Clean HTTP API for programmatic access. React-based dashboard for document management, search, and AI chat out of the box.

Interface

Architecture

A vertically integrated engine — from document ingestion to semantic retrieval to causal-chain audit — designed for PB-scale agent workloads.

Single Binary or PG Extension

Deploy as a standalone server, Docker container, or pgCortrix PostgreSQL extension — whatever fits your stack.

Content-Level Tracing

Others index call chains. Cortrix stores the actual content at every decision point — full causal-chain audit for AI compliance.

Universal Connectors

Native integration with LangChain, Dify, RAGFlow, MCP, and CLI. Plus HTTP upload, filesystem watchers, and CDC connectors.

Get Started in Minutes

From zero to semantic search in three commands.

1

Pull & Run

Terminal

docker pull cortrix/cortrix:latest
docker run -d -p 8080:8080 --name cortrix cortrix/cortrix:latest

2

Upload Documents

Terminal

curl -X POST http://localhost:8080/api/v1/documents/upload \
  -F "file=@your-document.pdf" \
  -F "namespace=default"

3

Semantic Search

Terminal

curl -X POST http://localhost:8080/api/v1/query \
  -H "Content-Type: application/json" \
  -d '{"query": "How does authentication work?", "namespace": "default"}'

Or build from source:

Terminal

git clone https://github.com/cortrix/cortrix.git
cd cortrix && mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j$(nproc)
./cortrix-server --port 8080

Built for the Agents That Matter

Cortrix is the data backbone for autonomous AI agents — not coding assistants, but agents that run real business processes.

Autonomous AI Agents (OpenClaw & Beyond)

The next wave of AI agents — like OpenClaw — need persistent semantic memory and auditable decision trails. Cortrix provides both natively.

AI Employee Audit & Compliance

Full causal-chain traceability for AI workers. Every decision, every data source, every reasoning step — stored and retrievable at content level, not just index level.

Workflow Orchestration

Integrate with LangChain, Dify, RAGFlow, and any MCP/CLI-based workflow. Cortrix acts as the semantic layer your orchestrator reads and writes to.

Enterprise Knowledge Infrastructure

From PostgreSQL (pgCortrix) to standalone engine — deploy semantic storage wherever your data lives. CDC connectors keep everything in sync.

How Cortrix Compares

A unified engine vs. assembling pieces.

Capability	Cortrix	Vector DB + RAG Framework
Document Ingestion	Built-in SPC pipeline	Separate parser + chunker
Embedding	Built-in (bge-m3, ONNX)	External API call
Vector Search	P-HNSW, in-process	Separate vector DB
Keyword Search	FTS5 + BM25	Often missing or separate
Hybrid Fusion	RRF built-in	Custom glue code
AI Memory	Native session memory	Not included
Causal-Chain Audit	Content-level tracing	Index-level only (if any)
PostgreSQL Integration	pgCortrix extension	Separate service
Workflow Connectors	LangChain / Dify / RAGFlow / MCP	Framework-specific
MCP Server	Built-in	Not available
Scale Target	PB ~ EB native	GB ~ TB typical
Deployment	Single binary / Docker / PG ext	Multiple services

Join the Community

Cortrix is open source and community-driven. We welcome contributions of all kinds.

Star on GitHub Join Discord

Report Issues Contributing Guide AGPL-3.0 License