Agent-Native Semantic Storage — ingest, understand, remember, and audit.
One unified engine from document to decision, with hybrid retrieval, reranking, and full causal-chain traceability.
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 ingestion, hybrid retrieval with reranking, interaction memory, and causal-chain audit into a single engine with REST, MCP, and SDK interfaces.
Semantic Processing Chain (SPC)
Ingest any document — PDF, Word, Markdown — through an automated pipeline: Docling parsing with OCR fallback, parent-child chunking, NER + summary enrichment, embedding, and indexing in one step.
Core
Hybrid Query + Reranker
Vector similarity (P-HNSW with WAL persistence) + BM25 keyword search fused via RRF, then precision-reranked with bge-reranker-v2-m3 — for best-in-class retrieval accuracy.
Core
AI Interaction Memory
Persistent conversation memory with session management, LLM-based fact extraction, typed memory (fact / preference / event) with decay, per-user isolation, transparency APIs, and full audit trail.
Core
Agent Observability & Self-Learning
Session, trace, and agent headers flow end-to-end. Retrieval feedback captures which chunks actually helped — every chunk accumulates a useful_ratio score that quietly reranks future results. Your storage gets smarter with every query, no retraining required.
Core
Causal-Chain Traceability
Open-source content-level provenance, not just call indexes. Every answer links back to the exact chunks, documents, and conversation turns that produced it — built-in citation tracing, retrieval attribution, and reasoning chain for debugging and trust.
Core
Namespaces + Cross-NS Query
Isolate data by project, team, or tenant with independent storage and indexes. Query across any subset of namespaces in parallel via scatter-gather, with reranker-unified ranking.
Core
Advanced RAG Techniques
Parent-Child chunking for precision + context, Contextual Retrieval for chunk disambiguation, RAG-Fusion multi-query, CRAG retrieval grading, and HyPE hypothetical-question indexing — built in, not bolted on.
Core
pgCortrix Extension
Semantic storage as a PostgreSQL extension. Bring agent-native capabilities directly into your existing Postgres infrastructure — no separate service needed.
Integration
MCP Server
Model Context Protocol server exposes Cortrix capabilities to Claude Code, Cursor, and any MCP-compatible AI tool out of the box.
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, DB import, and custom data sources.
Ecosystem
Python SDK + REST + Web UI
pip install cortrix for type-safe Python clients, a clean HTTP API for any language, and a built-in dashboard for document management, search, and AI chat.
Interface
Architecture
A vertically integrated engine — from document ingestion to hybrid retrieval with reranking to causal-chain audit — designed for PB-scale agent workloads.
Docker or PG Extension
Deploy via docker-compose one-command stack, standalone server, or pgcortrix PostgreSQL extension — whatever fits your stack.
Learns From Every Query
Retrieval feedback closes the loop: each chunk accumulates a useful_ratio based on what actually helped the agent, quietly lifting relevance over time — no fine-tuning, no extra pipeline.
Universal Connectors
Native integration with LangChain, Dify, RAGFlow, MCP, and CLI. Plus HTTP upload and filesystem watchers.
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 use the Python SDK:
Python
pip install cortrix
from cortrix import Client
client = Client("http://localhost:8080")
client.documents.upload("your-document.pdf", namespace="default")
results = client.query("How does authentication work?", namespace="default")
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
Reranker
bge-reranker-v2-m3 built-in
External service or missing
Cross-Namespace Query
Scatter-gather + unified ranking
Client-side orchestration
Advanced RAG
Parent-Child / Contextual / CRAG / HyPE
DIY or framework plugins
AI Memory
Typed memory with decay + LLM extraction
Not included
Causal-Chain Traceability
Content-level citation tracing
Index-level only (if any)
Agent Observability
Session / trace / feedback signals
Not included
Self-Learning Retrieval
useful_ratio feedback · no retraining
Not included
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.
