ToolsRecipesCompareAboutSponsor

Local tools for asking questions of your own documents

RAGShelf compares 34 self-hosted and desktop tools that let a local model answer questions from your PDFs, Office files, scans, and notes, without sending them to the cloud. Filter by license, hardware, document type, and whether it can run fully offline.

Browse setup recipes  Submit a tool

ToolLicenseHardwareDocumentsFully offlineLast updated
AnythingLLM
Active		MITDesktop or Docker; built-in engine runs CPU or GPU (NVIDIA CUDA, AMD/Intel Vulkan)PDF, DOCX, TXT, CSV, web pages, audio✅ Yes2026-10-24
Cherry Studio
Active		AGPL-3.0Desktop app (Windows, macOS, Linux)PDF, Office docs, images, text (OCR options)✅ Yes2026-10-24
Dify
Active		Dify Open Source License (Apache-2.0 + conditions)Docker; minimum 2 CPU cores, 4GB RAMPDF, DOCX, XLSX, HTML, MD✅ Yes2026-10-24
Docling
Active		MITCPU OK, GPU fasterPDF, DOCX, PPTX, XLSX, HTML, EPUB, images, audio, email (EML/MSG), LaTeX✅ Yes2026-10-24
Flowise
Active		Apache-2.0 (enterprise dir separate)Node serverAny via loaders✅ Yes2026-10-15
GPT4All (LocalDocs)
Stale (no release in 12+ months)		MITDesktop; CPU OKPDF, TXT, MD, DOCX folders✅ Yes2026-10-24
h2oGPT
Stale (no release in 12+ months)		Apache-2.0GPU recommendedPDF, images, audio, DOCX, XLSX✅ Yes2026-10-15
Haystack
Active		Apache-2.0FrameworkAny via converters✅ Yes2026-10-15
Jan
Active		Apache-2.0 per repo (GitHub reports NOASSERTION)Desktop; CPU/GPU (Vulkan, Metal, CUDA, ROCm). macOS: 8GB RAM for 3B, 16GB for 7B, 32GB for 13B modelsFile attachments✅ Yes2026-10-24
Khoj
Active		AGPL-3.0Server; GPU optionalPDF, Markdown, Org, Word, Notion, images✅ Yes2026-10-24
KoboldCpp
Active		AGPL-3.0Single binary; CPU/GPUText DB / file upload✅ Yes2026-10-15
Kotaemon
Active		Apache-2.0Docker; GPU optionalPDF, HTML, MHTML, XLSX in lite image; DOCX and more need the full image (Unstructured)✅ Yes2026-10-24
Langflow
Active		MITPython serverAny via components✅ Yes2026-10-15
LibreChat
No tracked release		MITServer + separate RAG API with pgvectorPDF, DOCX, TXT⚠️ Needs config2026-10-15
LightRAG
Active		MITDepends on chosen LLM; graph extraction wants a 14B+ modelText, PDF, DOCX; images/tables/Office via RAG-Anything✅ Yes2026-10-24
LlamaIndex
Active		MITFrameworkAny via readers✅ Yes2026-10-15
LocalAI
Active		MITCPU/GPUBackend (LLM, embeddings, rerank)✅ Yes2026-10-15
LocalGPT
No tracked release		MITGPU recommendedPDF, TXT, DOCX✅ Yes2026-10-15
Microsoft GraphRAG
Active		MITIndexing is LLM-heavyText, CSV✅ Yes2026-10-15
Msty Studio
No tracked release		Proprietary (free tier; Aurum $149/yr; lifetime $349)Desktop appPDF, DOCX, TXT, folders, YouTube⚠️ Needs config2026-10-15
Ollama
Active		MITCPU/GPU/AppleBackend✅ Yes2026-10-15
Onyx
Active		MIT core + enterprise (GitHub NOASSERTION)Docker; 16GB RAM suggested40+ connectors, PDFs, Office✅ Yes2026-10-15
Open Notebook
Active		MITDockerPDF, web, audio, video✅ Yes2026-10-15
Open WebUI
Active		Custom (BSD-3 based, branding clause)Light server; LLM via OllamaPDF, DOCX, TXT, web✅ Yes2026-10-24
paperless-gpt
Active		MITNeeds Paperless-ngx + LLMScanned docs✅ Yes2026-10-15
Paperless-ngx
Active		GPL-3.0Low (Raspberry Pi OK)Scans, PDFs, images, email✅ Yes2026-10-15
PipesHub
Active		Apache-2.0Docker stackOffice docs, connectors✅ Yes2026-10-15
PrivateGPT
Active		Apache-2.08GB+ RAM; GPU recommendedPDF, DOCX, TXT, CSV (tabular via tools), EPUB✅ Yes2026-10-24
Quivr
Stale (no release in 12+ months)		MITPython libraryPDF, TXT, MD, DOCX⚠️ Needs config2026-10-15
R2R
Stale (no release in 12+ months)		MITDocker, PostgresPDF, DOCX, images, audio⚠️ Needs config2026-10-15
RAG Web UI
Active		Apache-2.0DockerPDF, DOCX, MD, TXT✅ Yes2026-10-15
RAGFlow
Active		Apache-2.04 cores, 16GB RAM, 50GB disk recommended (Docker)PDF, scans, DOCX, XLSX, PPT, images✅ Yes2026-10-24
Unstructured (OSS)
Active		Apache-2.0CPU; hi_res needs more25+ formats incl. scans, email✅ Yes2026-10-15
Verba
Stale (no release in 12+ months)		BSD-3-ClauseDocker + WeaviatePDF, DOCX, TXT, CSV✅ Yes2026-10-15

"Fully offline: yes" means it can run with no cloud calls once configured with local models. "Needs config" means a default (e.g. OpenAI embeddings or a cloud reranker) must be swapped out. Stale = no release in over 12 months.

The RAGShelf weekly

New local RAG tools, releases, and tested setups. One email a week, no spam.