The best open-source RAG tools in 2026
The best open-source RAG tools right now are Qdrant, pgvector and Docling, out of 27 we have reviewed. Each is ranked by our rating, with its price, our verdict and who it suits.
Updated Oct 23, 2026 · Ranked from our reviews · How we review
- 1
Filtered vector search done properly makes Qdrant the open-source database to beat for production RAG, as long as someone locks down a server that ships open.
Best for: Teams running production RAG or search that need fast filtered vector queries.
Price: Free and open source under the Apache 2.0 licence. Qdrant Cloud has a free 1 GB cluster, then usage-based billing; Premium, Hybrid and Private Cloud are priced by sales.
Read our Qdrant review4.7/5 - 2
For most teams already on Postgres, a dedicated vector database is extra infrastructure that pgvector makes unnecessary.
Best for: Teams already on Postgres building RAG or semantic search over a few million vectors.
Price: Free and open source under the PostgreSQL License. The only cost is the Postgres instance you already run, or a larger one.
Read our pgvector review4.6/5 - 3
Docling is the open-source parser that gets tables right, and the one whose security releases you cannot afford to skip.
Best for: Teams building self-hosted RAG over PDFs and Office documents with complex tables.
Price: Free and open source under the MIT license.
Read our Docling review4.5/5 - 4
Built for billions of vectors and priced in operational effort; below a few hundred million, it is more database than most teams need.
Best for: Platform teams running vector search at hundreds of millions to billions of vectors.
Price: Free and open source under Apache 2.0. Zilliz Cloud is the managed service, with a free tier; published paid prices vary by source, so use Zilliz's calculator.
Read our Milvus review4.4/5 - 5
For agents whose facts change over time, Graphiti is the open-source memory layer to beat, provided you will run a graph database and pay for LLM extraction on every write.
Best for: Teams building long-running agents that must remember changing facts about users, accounts or the world.
Price: Free and open source under Apache 2.0. You supply the graph database and pay for LLM and embedding calls during ingestion.
Read our Graphiti review4.3/5 - 6
Still a deep toolkit for document-heavy RAG, but its maker now says plainly that the framework is no longer its priority.
Best for: Python developers building RAG or document agents over large, varied data sources.
Price: Framework free under MIT. LlamaParse (paid cloud): Free with 10k credits, Starter $50/month (40k credits), Pro $500/month (400k credits), Enterprise custom; 1,000 credits = $1.25.
Read our LlamaIndex review4.3/5 - 7
Marker turns hard PDFs into usable Markdown better than any other open pipeline its makers measured, though the weights licence puts a revenue ceiling on that freedom.
Best for: RAG and data teams converting large volumes of complex or scanned PDFs on their own hardware.
Price: Code free under Apache 2.0; weights free below $5M revenue and funding, commercial licence above. Hosted Datalab API: Convert $4 per 1,000 pages (fast or balanced); Team plan $400/month.
Read our Marker review4.3/5 - 8
Repomix packs a small repo into one prompt better than anything else, and becomes the wrong tool the moment the repo outgrows the context window.
Best for: Developers who want to hand a small or medium repo to a chat model in one file.
Price: Free and open source (MIT). The web version at repomix.com is free; there is no paid tier.
Read our Repomix review4.3/5 - 9
Still the strongest open crawler for LLM pipelines, but its Docker server's 2026 security record rules out exposing it without hardening.
Best for: Python teams that want a self-hosted crawler for RAG or agents without per-page costs.
Price: Library and Docker server free (Apache 2.0 plus an attribution clause). Crawl4AI Cloud: 1,000 free credits a month; Starter $10 a month for 10,000 credits.
Read our Crawl4AI review4.2/5 - 10
A hosted web layer for agents that earns its fee; the open-source edition is a starting point, not a substitute.
Best for: Developers who want a managed API to turn websites into LLM-ready data.
Price: AGPL-3.0 open source. Cloud: Free 1,000 credits/month; Hobby $19, Standard $99, Growth $399, Scale $749 a month; Enterprise custom.
Read our Firecrawl review4.2/5 - 11
A framework for people who want to read their RAG system rather than trust it, at the price of writing more of the wiring themselves.
Best for: Engineering teams building maintainable production RAG and agent pipelines in Python.
Price: Free and open source under the Apache 2.0 licence. deepset Studio is free for one user; the Haystack Enterprise Platform is custom-priced.
Read our Haystack review4.2/5 - 12
Hybrid keyword-and-vector search is still Weaviate's best argument, but its newest operational features now sit behind a licence key.
Best for: Teams building search or RAG where exact keywords and semantic similarity both matter.
Price: Core is BSD-3-Clause; some enterprise features need a paid licence key. Weaviate Cloud: Free (100,000 objects), Flex pay-as-you-go with a $45/month minimum, Premium from $400/month on a prepaid contract.
Read our Weaviate review4.2/5 - 13
Quick to embed and cheap to host for read-heavy RAG, though nothing has been tagged since May 2026.
Best for: Developers prototyping RAG or running small to mid-size vector search.
Price: Free and open source under Apache 2.0. Chroma Cloud: Starter $0 plus usage with $5 credits; Team $250/month plus usage; Enterprise custom. Writes $2.50/GiB, storage $0.33/GiB-month.
Read our Chroma review4.1/5 - 14
The default graph store when your RAG or agent memory genuinely needs relationships, but the free edition stops at one server and most RAG does not need a graph at all.
Best for: Teams whose RAG or agent-memory questions depend on relationships between entities, and who need a proven graph store.
Price: Community Edition free under GPLv3. Enterprise by quote. AuraDB: Free tier; Professional $0.09/GB/hour (1 GB minimum); Business Critical $0.20/GB/hour (2 GB minimum).
Read our Neo4j review4.1/5 - 15
When your embeddings live next to images, video or training data, LanceDB removes the separate database entirely, though busy multi-user search is not what the free library is built for.
Best for: Teams working with multimodal or training data who want vector search without running a server.
Price: Free and open source under Apache 2.0. LanceDB Enterprise is sold on annual commitments; AWS Marketplace lists a $60,000 12-month commitment, with private offers by quote.
Read our LanceDB review4.0/5 - 16
Mem0 is still the quickest way to give an agent a memory, but the open-source SDK is increasingly the trial version of a platform that keeps graph memory and the best benchmark scores for itself.
Best for: Developers adding cross-session user memory to chatbots and agents.
Price: Open-source SDK free under Apache 2.0. Platform: Hobby free (10k adds, 1k retrievals/month); Starter $19/month; Pro $249/month (500k adds, 50k retrievals, graph memory); Enterprise custom.
Read our Mem0 review4.0/5 - 17
A finished document-to-answer product you can run on one server, whose security record says it belongs behind a VPN rather than on the open internet.
Best for: Teams that want a self-hosted document Q&A and knowledge assistant without building a pipeline.
Price: Free and open source under the Apache 2.0 licence. You pay for the server and the model calls.
Read our RAGFlow review4.0/5 - 18
Airbyte has the widest open catalogue for pulling business data into a RAG index, but its AI destinations lag its ELT core and self-hosting it means owning its security patches.
Best for: Data teams feeding a RAG system from many SaaS apps and databases who want incremental syncs they can self-host.
Price: Self-hosted Core is free (MIT and ELv2). Cloud Standard from $20/month with 5 credits; Plus from $189/month for 40 credits up to $4,999 for 2,000; Pro and Enterprise Flex custom.
Read our Airbyte review3.9/5 - 19
The sensible way to try graph RAG, cheaper to maintain than Microsoft GraphRAG, but most teams should prove they need a graph before adding one.
Best for: Teams whose evaluations show multi-hop questions failing with hybrid search and who want a graph on Postgres.
Price: Free and open source under the MIT license. Indexing and queries use your own LLM and embedding credits.
Read our LightRAG review3.9/5 - 20
code2prompt is the packer for people who care about the prompt template, held back by a release trail that no longer lines up.
Best for: Terminal users who want templated prompts or git-diff prompts built from their code.
Price: Free and open source (MIT).
Read our code2prompt review3.8/5 - 21
Cognee builds a shared knowledge graph from almost any source, but its defaults need an owner before they meet production.
Best for: Teams building several agents that need a shared knowledge graph assembled from many sources.
Price: Open source under Apache 2.0. Cognee Cloud: free up to 1M tokens and one workspace, then $1 per 1M tokens plus $5 per extra workspace a month; Enterprise is a quoted BYOC engagement.
Read our Cognee review3.8/5 - 22
MarkItDown gets Office files into a model faster than anything else open source, and flattens the PDFs that retrieval most needs kept intact.
Best for: Developers who need to feed Office documents and clean digital PDFs to a model quickly and cheaply.
Price: Free and open source under the MIT licence. Optional Azure conversion services are billed by Microsoft.
Read our MarkItDown review3.8/5 - 23
Still the widest net for odd file formats, but its makers now publish the numbers showing why you would pay them instead.
Best for: Teams building ETL pipelines that ingest many document formats for RAG.
Price: Library free under Apache 2.0. Hosted platform: 10,000 free pages, then $0.015 a page; Business plan with VPC or dedicated deployment priced on request.
Read our Unstructured review3.8/5 - 24
A tidy chunking library mid-rewrite: v1.7 works today, but most of what makes it distinctive is scheduled for removal in v2.
Best for: Teams that need a light recursive or code chunking step and can accept a v2 migration soon.
Price: Free and open source under the MIT licence.
Read our Chonkie review3.7/5 - 25
grepai trims an agent's code-reading bill without sending code anywhere, though its own benchmark puts the saving nearer a quarter than the 97% headline.
Best for: Developers who want their coding agent to search a repo by meaning, locally, instead of reading whole files.
Price: Free and open source (MIT). Runs locally with Ollama or LM Studio; OpenAI embeddings are billed by OpenAI.
Read our grepai review3.7/5 - 26
Good at pulling a few typed fields from messy pages, but it ships your scraped content to its makers by default and pays a model call for every page.
Best for: Developers extracting typed fields from a modest number of varied web pages.
Price: Library free under MIT; you pay for LLM tokens. Hosted API: 500 one-off free credits, then $20, $100 or $500 a month; Enterprise custom.
Read our ScrapeGraphAI review3.6/5 - 27
Still the reference design for whole-corpus questions, but Microsoft has stopped developing it and the indexing bill is as large as ever.
Best for: Teams with a fixed, high-value document set where cross-document questions justify a costly one-off index.
Price: Free and open source under the MIT licence. Indexing runs on your own OpenAI or Azure OpenAI credits, which the README warns can be expensive.
Read our GraphRAG review3.5/5
More guides
- Best open-source tools to run LLMs locally
- Best open-source AI agent frameworks
- Best open-source LLM fine-tuning tools
- Best open-source AI chat apps
- Best open-source AI coding agents
- Best open-source LLM evaluation tools
- Best open-source LLM gateways and observability tools
- Best Claude Code plugins and agent skills