Compare RAG boilerplates
Verified RAG vs LangGraph RAG Agent
Compare the Verified RAG Delivery Kit with the LangGraph RAG Agent editions: evaluation and acceptance reports, vector database, agent design, databases and licensing.
BoilerShelf sells two different kinds of RAG boilerplate. The Verified RAG Delivery Kit is built to prove to a client that answers are correct. The LangGraph RAG Agent is a self-correcting agent that retrieves, grades and rewrites before it answers.
They are different codebases with different strengths, so the right one depends on what you have to deliver.
Side by side
At a glance
| Verified RAG | LangGraph RAG (MongoDB) | LangGraph RAG (PostgreSQL) | |
|---|---|---|---|
| Editions and prices | Developer $179 · Agency $499 | Basic $59 · Pro $99 | Basic $59 · Pro $99 |
| How it answers | Deterministic retrieval and generation with citation-grounded answers; not agentic | Self-correcting LangGraph loop: retrieve, grade, rewrite, generate | Self-correcting LangGraph loop: retrieve, grade, rewrite, generate |
| Proving answer quality | Batch evaluation test suites against golden answers, groundedness scoring and branded PDF acceptance reports | Not part of this product | Not part of this product |
| Vector database | PGVector (PostgreSQL) | Qdrant | Qdrant |
| Application database | PostgreSQL with Alembic migrations | MongoDB with LangGraph checkpointing | PostgreSQL with SQLAlchemy and Alembic migrations |
| LLM providers | OpenAI, Google Gemini, Anthropic Claude | Google Gemini, OpenAI, Groq | Google Gemini, OpenAI, Groq |
| Users and teams | Developer is single-user. Agency is multi-tenant: Super Admin, Organization Admin and Member roles, organization invitations and per-organization quotas | Basic has no login. Pro adds JWT authentication and an admin dashboard | Basic has no login. Pro adds JWT authentication and an admin dashboard |
| Voice | Not included | Pro: local Whisper speech-to-text and Kokoro text-to-speech | Pro: local Whisper speech-to-text and Kokoro text-to-speech |
| Client licensing | Developer: single project. Agency: multiple client projects with source handover | Single project | Single project |
| Difficulty | Advanced | Intermediate | Intermediate |
| Setup time | About 30 minutes | About 15 minutes | About 15 minutes |
| Deployment targets | Docker, Render, Railway, AWS | Docker, Render, Railway, AWS, DigitalOcean | Docker, Render, Railway, AWS, DigitalOcean |
Which one
How to choose
Choose Verified RAG if you deliver RAG to clients
You need to show that answers are grounded in the client's documents. It gives you citations, evaluation test suites you can re-run, and a branded acceptance report to hand over. The Agency edition is the one for serving several client organizations from one deployment.
View Verified RAG Delivery Kit · Developer $179 · Agency $499
Choose LangGraph RAG on MongoDB for a self-correcting agent
You are building your own AI product and want an agent that grades what it retrieved and rewrites the query when the result is weak. Pick the MongoDB edition if your application already runs on MongoDB.
View LangGraph RAG Agent — MongoDB Edition · Basic $59 · Pro $99
Choose LangGraph RAG on PostgreSQL for a relational stack
The same agent design for teams on PostgreSQL, with SQLAlchemy models and Alembic migrations. Pro adds login, an admin dashboard and a local voice pipeline.
View LangGraph RAG Agent — PostgreSQL Edition · Basic $59 · Pro $99
pgvector or Qdrant?
The Verified RAG Delivery Kit keeps documents, metadata and embeddings in one PostgreSQL database using the pgvector extension, and combines vector similarity with PostgreSQL full-text search. That means one fewer service to run and a single database to back up.
The LangGraph RAG Agent uses Qdrant, a dedicated vector database, next to your application database. It adds a service to run, and in return gives you a purpose-built vector store.
Neither choice is better in general. Choose the one that matches the infrastructure you already operate.
What none of the RAG kits promise
These are starting points you host yourself, not hosted services. You bring your own LLM provider key. Verified RAG ships without an automated test suite or CI/CD pipeline, and its citation checks show that an answer cites a source, not that the answer is factually correct. Each product page lists exactly what is and is not included.