Skip to content

Hjarni vs Ragie

Ragie hosts a RAG pipeline for documents you upload. Hjarni is a knowledge base you write in directly. Different shape, different audience.

Free to start. No credit card required.

Quick answer

Ragie is a managed RAG pipeline: you upload files (PDFs, Word, Markdown) and Ragie handles chunking, embeddings, and reranking, exposed through a hosted MCP server. Hjarni is a Markdown knowledge base you write and edit in directly, with a built-in MCP server and full-text search, so there is no ingestion step. Pick Ragie if your source of truth is a large corpus of documents you ingest at scale, or you are building a RAG feature on top. Pick Hjarni if your source of truth is the note you write, with folder-level AI instructions across ChatGPT, Claude, and Cursor.

Hjarni Ragie
Primary surface

Hjarni is a knowledge base. Ragie's primary input is files (PDFs, Word, Markdown) you hand off to a managed pipeline.

Markdown notes you write Documents you upload
MCP server

Both expose a hosted MCP server. Hjarni's is part of the knowledge base. Ragie's is a remote endpoint over the hosted retrieval index.

Built-in Built-in
Bring your own AI

Both are designed for ChatGPT, Claude, Cursor, and other MCP clients.

Edit content in the product

Hjarni notes are editable in the app. Ragie is built around documents you maintain elsewhere and ingest.

Re-upload to update
Folder structure for humans

Hjarni uses nested folders and tags. Ragie uses partitions and document metadata for scoping.

Partitions and metadata
Custom AI instructions per folder

Hjarni lets you set folder-level rules an AI must follow. Ragie does not have an equivalent.

Built-in
Team collaboration

Both support multi-user setups. Hjarni's team plan is folder-based; Ragie's is partition-based.

Reranking and retrieval tuning

Ragie ships reranking, hybrid retrieval, and other RAG knobs. Hjarni's MCP server uses full-text search for now.

Full-text search Built-in
Pricing model

Hjarni is flat-priced. Ragie meters usage: a monthly tier, plus per-page storage charges past the included pages and per-minute charges for audio and video, so the bill climbs as the corpus grows.

Flat plans with note quotas Metered: monthly tiers plus per-page and per-minute overage
Free tier
Best fit

If the source of truth is the document, Ragie. If the source of truth is the note, Hjarni.

Notes you write and edit Files you ingest at scale

A knowledge base vs a RAG pipeline

Ragie's job is to take a pile of documents and turn them into a context-aware API. You upload PDFs, Word docs, and Markdown. Ragie handles chunking, embeddings, hybrid retrieval, and reranking. Your AI calls the hosted MCP endpoint and gets relevant snippets back. The hard parts of building a RAG system are abstracted away.

Hjarni is a different product. You write notes in the app. The notes are Markdown, organized in folders, tagged, linked. The built-in MCP server lets ChatGPT, Claude, and other clients search and read them. There is no upload step because the notes already live in the app.

Different inputs, different workflows

With Ragie, the document is the source of truth. You write a brief in Word, upload it, and any change goes through re-uploading or syncing. The team's writing tools stay where they are.

With Hjarni, the note is the source of truth. You write in the app, edit in the app, and the next time your AI asks, it sees the edit. There is no ingestion lag because there is no ingestion.

Ragie indexes the documents you already have. Hjarni is the place you write the notes in the first place.

Infrastructure you build on vs a product you use

Ragie is developer infrastructure. It ships an SDK and connectors, and the assumption is that you build an application on top of the retrieval API. That is the right shape if you are shipping a RAG feature inside your own product, but it means there is a pipeline to wire up before anyone gets an answer. The pricing is metered to match: a monthly tier, plus per-page charges once storage runs past the included pages, plus per-minute charges for audio and video. The bill tracks volume, so a growing corpus is a growing invoice.

Hjarni is the opposite shape. It is ready to use the moment you connect a client, with the knowledge base and the MCP server in one product. There is no application to build on top, no SDK to learn, and the plans are flat with note quotas rather than per-page or per-minute metering. You write a note, your AI reads it, and the price does not move because you added more pages or a longer recording.

When Ragie is the better fit

If you already have hundreds or thousands of files, a steady ingestion pipeline matters, and you want to ship a RAG feature inside your own product, Ragie is built for that. Their value is in the pipeline you do not have to operate.

When people choose Hjarni instead

The switch usually happens when the workflow is "I want to write notes that my AI can read", not "I have a corpus of files I want to query". Hjarni is the knowledge base and the MCP server in one product. There is no document pipeline to feed, no embedding model to pick, no chunking strategy to tune. You write a note, your AI reads it.

Per-folder AI instructions also matter. Tell Claude to keep one folder formal and another casual. Tell Cursor to skip the marketing folder when answering coding questions. Ragie does not have an equivalent because its unit is a document, not a folder.

When to use Ragie

  • You already have a large corpus of files
  • You want a managed RAG pipeline with reranking
  • You are building a feature on top, not using it as a knowledge base

When to use Hjarni

  • You want a knowledge base with a built-in MCP server
  • You want to write in the product, not upload to it
  • You want folder-level AI instructions across ChatGPT, Claude, and Cursor

Ragie hosts the pipeline. Hjarni hosts the notes.

Common questions

Common questions

What is Ragie?

A hosted RAG-as-a-service. You upload PDFs, Word, or Markdown files; Ragie handles chunking, embeddings, hybrid retrieval, and reranking, and exposes the result through a hosted MCP server.

Is Hjarni a RAG product?

Hjarni's MCP server uses full-text search over notes you write in the app. Whether you call that RAG depends on how strictly you define the term. The end-to-end behaviour is the same shape: the AI retrieves before it answers.

Can I migrate from Ragie to Hjarni?

If the source files are Markdown, drop the ZIP into Hjarni's importer and the folder structure comes across. PDFs and Word docs are not first-class in Hjarni today, so plan to convert or summarize them into notes first.

Which is cheaper?

It depends on volume. Hjarni's free tier includes full MCP access. Ragie's pricing is metered on pages indexed and retrievals served. For a notes-style workload, Hjarni is usually cheaper. For a thousand-file ingestion pipeline, Ragie is built for that.

Can I use both?

Yes. Some teams keep Ragie for product-side document retrieval and use Hjarni as the place humans actually write decisions and runbooks. The AI client can connect to both MCP servers in parallel.

How does Ragie's pricing compare to Hjarni's, and do I have to build something on top of it?

Ragie is RAG-as-a-Service: it ships an SDK and connectors and expects you to build an application on top of its retrieval API, and it prices by usage. You pay a monthly tier plus per-page storage charges once you pass the included pages and per-minute charges for audio and video, so the bill grows with your corpus. That is the right tool if you are shipping a RAG feature inside your own product. Hjarni is the opposite shape: a ready-to-use Markdown knowledge base with a built-in MCP server that ChatGPT, Claude, and other clients read and write directly. There is no pipeline to wire up and no SDK to learn, and the plans are flat with note quotas rather than per-page or per-minute metering, so adding more pages or a longer recording does not move the price.

Start here

Write once. You both remember.

Free to start. No credit card required.

Works with Claude and ChatGPT today.