Updated
What makers build

RAG & AI memory tech stack 2026: what 50 real products use

Retrieval engines, knowledge-graph RAG and memory layers for AI apps.

Among the 50 products we track in RAG & AI memory, the usual stack is OpenAI API (LLM API, 83%), PostgreSQL (database, 66%), React (frontend framework, 56%) and pgvector (vector database, 50%): shares of the products that show each decision.

How they're built

The typical stack

The leading pick where at least 8 of them show the decision and the leader has at least a quarter of it.

DecisionMost common pickShareAt default usage
LLM APIOpenAI API34 of 41 · 83%$10/mo · GPT-6 Luna
DatabasePostgreSQL25 of 38 · 66%—
Frontend FrameworkReact20 of 36 · 56%—
Vector Databasepgvector16 of 32 · 50%—
Backend FrameworkFastAPI25 of 28 · 89%—
AI SDK & Agent FrameworkLangChain16 of 27 · 59%—
Document Parsingpypdf14 of 24 · 58%—
HostingVercel11 of 24 · 46%$40/mo · Pro + Flat Rate CDN 10M
MCP Server for Your ProductMCP SDK18 of 23 · 78%—
Database Access & ORMsSQLAlchemy16 of 17 · 94%—
EmbeddingsSentence Transformers11 of 16 · 69%—
LLM Observability & EvalsLangfuse7 of 14 · 50%$29/mo · Core
Open Model Hosting & GPUsHugging Face13 of 14 · 93%—
Transactional EmailResend10 of 13 · 77%$20/mo · Pro 50k
Run Models LocallyOllama10 of 11 · 91%—
Background Jobs & CronRabbitMQ4 of 9 · 44%—
File StorageMinIO7 of 9 · 78%—

Monthly bills add up to about $99 at the calculators' default usage, list prices. Set your own usage →

Where they differ from other products

Tools RAG & AI memory pick at least twice as often as products overall. Each bar is the share of products that show the decision and chose the tool.

What they chose, decision by decision

Among the RAG & AI memory that show each choice, from makers' products and open-source code alike. Under each tool, some of the products that chose it.

LLM API41 products
OpenAI API34 · 83%Vectorize, Byterover, Supavec
Claude17 · 41%Byterover, Weavable, Nia
LiteLLM16 · 39%puppyone-cloud, cocoindex, Memori
Gemini API10 · 24%agentset, Memori, semantica
Database38 products
PostgreSQL25 · 66%Byterover, puppyone-cloud, agentset
Neo4j14 · 37%Vectorize, Byterover, cocoindex
SQLite12 · 32%Memori, semantica, m_flow
MySQL10 · 26%moss, FastGPT, cocoindex
React20 · 56%puppyone-cloud, agentset, moss
Next.js19 · 53%Vectorize, Byterover, Cloudglue
Vue.js6 · 17%moss, KnowFlow, ragflow-plus
Streamlit5 · 14%semantica, ai-optimizer, Langchain-Chatchat
Vector Database32 products
pgvector16 · 50%puppyone-cloud, FastGPT, cocoindex
Qdrant11 · 34%Byterover, puppyone-cloud, cocoindex
Milvus10 · 31%puppyone-cloud, FastGPT, semantica
Faiss8 · 25%cocoindex, Memori, semantica
FastAPI25 · 89%puppyone-cloud, moss, Acontext
Flask5 · 18%puppyone-cloud, moss, KnowFlow
Express4 · 14%FastGPT, MemOS, embedbase
Hono2 · 7%FastGPT, MemOS
LangChain16 · 59%Nia, IgnitionRAG, Memori
LlamaIndex8 · 30%agentset, moss, FastGPT
LangGraph5 · 19%MemOS, MemoryBear, Yuxi
Instructor5 · 19%puppyone-cloud, cocoindex, semantica
Document Parsing24 products
pypdf14 · 58%PageIndex, m_flow, ai-optimizer
MarkItDown7 · 29%SAG, MemOS, ApeRAG
Docling4 · 17%semantica, ai-optimizer, trustgraph
PyMuPDF4 · 17%puppyone-cloud, MemoryBear, Langchain-Chatchat
Hosting24 products
Vercel11 · 46%Vectorize, Cloudglue, Nia
GitHub Pages7 · 29%semantica, ai-optimizer, Yuxi
Cloudflare3 · 13%Acontext, Supametas.AI, honcho
Fly.io3 · 13%Cloudglue, semantica, honcho
MCP SDK18 · 78%moss, FastGPT, Acontext
FastMCP8 · 35%puppyone-cloud, ai-optimizer, redis-vl-python
SQLAlchemy16 · 94%puppyone-cloud, Acontext, Memori
Embeddings16 products
Sentence Transformers11 · 69%puppyone-cloud, cocoindex, semantica
Cohere7 · 44%puppyone-cloud, agentset, redis-vl-python
Voyage AI6 · 38%agentset, redis-vl-python, KnowFlow
Langfuse7 · 50%FastGPT, KnowFlow, ragflow-plus
LangSmith7 · 50%semantica, m_flow, MemOS
Ragas2 · 14%LightRAG, ApeRAG
Hugging Face13 · 93%semantica, chonkie, MemOS
Fireworks AI1 · 7%Langchain-Chatchat
vLLM1 · 7%NeMo-Retriever
Resend10 · 77%Byterover, Cloudglue, Nia
SendGrid3 · 23%Vectorize, Memori, ai-optimizer
Ollama10 · 91%semantica, redis-vl-python, MemOS
RabbitMQ4 · 44%Acontext, semantica, MemOS
Celery3 · 33%semantica, ApeRAG, MemoryBear
File Storage9 products
MinIO7 · 78%puppyone-cloud, FastGPT, KnowFlow
Amazon S33 · 33%agentset, FastGPT, atomicmemory
UploadThing1 · 11%puppyone-cloud

Small samples, fewer than 8 products: AI UI Design (shadcn/ui 7) · Payments (Stripe 6) · Error Monitoring (Sentry 5) · Product Analytics (PostHog 4, Mixpanel 1) · Search APIs for AI (Tavily 5, SerpApi 2) · Web Analytics (Google Analytics 2, Plausible Analytics 1) · Web Scraping for AI (Crawl4AI 2, Firecrawl 1) · Authentication (Clerk 1, Better Auth 1) · Code Sandboxes for Agents (E2B 3) · Speech-to-Text & Text-to-Speech (ElevenLabs 1, Cartesia 1)

RAG & AI memory we track

24 makers' products and 26 open-source projects, makers' products first, each with the tools it uses (the most common here first).

Questions

What tech stack is behind RAG & AI memory?

Among the 50 products we track in RAG & AI memory, the usual stack is OpenAI API (LLM API, 83%), PostgreSQL (database, 66%), React (frontend framework, 56%) and pgvector (vector database, 50%): shares of the products that show each decision. After those, FastAPI leads backend framework, LangChain leads AI SDK & agent framework and pypdf leads document parsing.

Which tools are more common in RAG & AI memory than elsewhere?

Neo4j for database (13× as often as among products overall), Streamlit for frontend framework (5.5× as often as among products overall) and Qwen API for LLM API (5.1× as often as among products overall).

How much does a typical RAG & AI memory stack cost?

About $99 a month in bills at the pricing calculators' default usage and list prices. The biggest lines: Vercel $40, Langfuse $29 and Resend $20.

Which products in RAG & AI memory are open source?

26 of the 50 are open-source projects, known from their public code, among them redis-vl-python, chonkie, MemOS, KnowFlow, ragflow-plus and LightRAG.