Airgap
Local document AI: query your PDFs, Markdown and text files with page-level citations — nothing leaves your machine.
Airgap earns its place when the requirement is hard: privileged legal files, employee records or anything under a data-residency rule. The page-and-paragraph citations are the real selling point — they make local RAG auditable, not just private, and the source highlighting means you check the evidence rather than trust the summary. Local generation through llama.cpp with five built-in models (Llama 3.2, Qwen 2.5, Phi 3.5, Gemma 2) and any GGUF repo as a bring-your-own option keeps it flexible without a cloud dependency. It is single-machine by design, with no hosted account or cloud workspace, so teams wanting shared workspaces or cloud co-editing should look at hosted document AI instead.
Verified 3d ago · liveness 59/100 · cite: rightaichoice.com/tools/airgap
- Legal teams querying privileged PDFs without uploading them to a cloud service
- HR staff searching employee records on a machine that never sends files out
- Researchers under NDA or data-residency rules who need a citation for every answer
- Analysts working in air-gapped or offline environments across confidential reports
- Teams that need a shared cloud workspace or real-time co-editing
- Users who want document AI on a phone or tablet
- Laptops without the GPU or CPU headroom to run local models comfortably
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Airgap if you need a shared cloud workspace, real-time co-editing with colleagues, or document AI on a phone or tablet — it is deliberately a single-machine, single-user tool.
You supply the hardware: running local generation comfortably needs a machine with real GPU or CPU headroom, and a laptop that struggles with it means a hardware upgrade is the real cost.
Airgap's pricing fits teams whose volume aligns with the published tiers. Compare against the alternatives listed below for stage-specific value.
In short
Airgap — Local document AI: query your PDFs, Markdown and text files with page-level citations — nothing leaves your machine. Best for Legal teams querying privileged PDFs without uploading them to a cloud service, HR staff searching employee records on a machine that never sends files out, Researchers under NDA or data-residency rules who need a citation for every answer. Paid pricing.
What people actually say about Airgap — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
50 mentions across 4 sources (Hacker News, YouTube, Product Hunt, Lemmy) · researched Aug 24, 2026.
Average across the 4 sources that answered — each source counts once, not each post.
- +Truly local processing ensures data never leaves your device.
- +Offline capability makes it useful for air-gapped environments.
- +Ideal for regulated industries with strict compliance needs.
- +No data residency concerns, appealing to global enterprises.
- +Simple conversational interface for asking questions over documents.
- −Very few user reviews limit confidence in reliability.
- −Setup requires technical knowledge, despite beginner claim.
- −On-device AI may lag on large or multiple documents.
- −Pricing unclear; no free tier or trial mentioned.
- −No community forum or support channel visible.
- • Hardware requirements for running on-device models
- • No free trial or demo version available
Viability Score
How well maintained and how widely used is Airgap? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Local document RAG over PDF, Markdown and text files
- Page and paragraph citations linking answers to source
- Source highlighting on cited passages
- Portable, shareable vaults with a private per-vault vector index
- Fully on-device pipeline — nothing uploaded to be indexed
- Offline ingestion, retrieval and generation once models are cached
- Five built-in local models: Llama 3.2, Qwen 2.5, Phi 3.5, Gemma 2
- Bring your own GGUF model from any Hugging Face repository
- Multilingual retrieval with a 100-language embedding model
- Local inference via llama.cpp, models cached after one download
- No hosted account, cloud workspace or relay between app and files
- Vaults protected at rest by FileVault, BitLocker or LUKS disk encryption
- Available for macOS, Windows and Linux
- Optional connected-model mode (off by default) that sends questions to a hosted API
- Works with Wi-Fi disabled for air-gapped environments
About Airgap
Airgap is a local document AI and RAG tool for people who cannot hand confidential files to a hosted service. You add PDFs, Markdown and text files to a vault — a portable container holding a project's documents and its private vector index. Airgap parses them, retrieves the relevant chunks, runs a language model on-device through llama.cpp, and links every answer back to the source page and paragraph. The pitch is provenance as much as privacy: instead of trusting a summary, you get a citation you can open, with the cited passage highlighted. Five models ship built in — Llama 3.2, Qwen 2.5, Phi 3.5 and Gemma 2 — and you can point Airgap at any GGUF repository on Hugging Face when you want a different local model. A default 100-language embedding model means documents are searched in the language they were written in rather than translated first. There is no hosted account, cloud workspace or relay between the app and your files; vaults are ordinary files in the app's data directory, protected at rest by whatever disk encryption your machine already runs — FileVault, BitLocker or LUKS. Once models are downloaded, ingestion, retrieval and generation keep working with Wi-Fi disabled, which is what makes it viable in air-gapped or offline setups. An optional connected-model mode is off by default; only then do questions leave the machine. It runs on macOS, Windows and Linux. The trade-off is architectural, not incidental: this is a single-machine tool, so shared cloud workspaces, real-time co-editing and mobile access are not part of the picture. If your constraint is that files can't leave the laptop, Airgap is built precisely for that. If you need a team workspace, a hosted document AI will fit better.
Behind the Verdict
Airgap's central claim is architectural, not promotional: retrieval, generation and citation all happen on your machine, and nothing is uploaded to be indexed. The homepage describes a pipeline of parse → retrieve → generate → cite, with each answer linked back to the original source page and the cited passage highlighted. For anyone whose blocker is "I can't put these files in a cloud service," that is the whole product. The details that matter to a buyer evaluating this: Local model path. Five models ship built in — Llama 3.2, Qwen 2.5, Phi 3.5 and Gemma 2 — run on-device through llama.cpp and cached after one download. You can also point Airgap at any GGUF model from a Hugging Face repository, which means the model layer isn't a dead end if your needs change. Multilingual retrieval. Vaults embed with a 100-language model by default, so a French contract is searched in French rather than machine-translated into English first. For researchers working across languages this is a real quality difference, not a checkbox. Vaults as the unit of work. A vault is a portable, shareable container holding a project's documents plus its private vector index. That makes "hand someone the corpus" possible without a hosted workspace — although the tool itself remains single-machine. Security model. There is no hosted account, no service account and no hosted relay between the app and your documents. Vaults are ordinary files in the app's own data directory, so they inherit whatever disk encryption your machine already runs — FileVault, BitLocker or LUKS. Offline operation is the payoff: once models are cached, ingestion, retrieval and generation keep working with Wi-Fi disabled. That is what makes the air-gapped use case real rather than aspirational. Connected model mode. Optional and off by default. Point Airgap at a hosted API and only then do questions leave the machine. Worth knowing it exists; worth knowing it's opt-in. The honest constraints. This is a single-machine tool, so shared cloud workspaces and real-time co-editing are not part of the picture, and neither is phone or tablet access. Local generation depends on your hardware — a laptop without GPU or CPU headroom will feel it — and models must be downloaded and cached before offline use works at all. The first-run experience therefore includes a download, not just an install. If your workflow assumes a browser tab and a shared link, Airgap is the wrong shape; if your workflow assumes the file never leaves the laptop, it is exactly the right one.
Researching Airgap? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Airgap actually fits — and what changes day-one when you adopt it.
Add the contract PDFs to a vault on your laptop, let Airgap build the private vector index, then ask which agreements cap liability at a specific figure. Each answer links back to the page and paragraph and highlights the cited passage, so you can open the clause and check it yourself.
Outcome: Key clauses located and verified against the source page without a single file being uploaded to a service.
Ingest a folder of papers and notes as a vault, then query it in the language the documents were written in — the default 100-language embedding model searches them as written rather than translating first.
Outcome: A literature summary where every claim carries a page-level citation you can follow back to the source.
Download and cache the models while you still have a connection, then disable Wi-Fi and keep working: ingestion, retrieval and generation all continue locally through llama.cpp.
Outcome: Confidential reports analysed on a machine that is physically incapable of sending the documents anywhere.
Use Cases
- Chat with privileged legal contracts to extract key clauses without uploading them
- Summarize lengthy research papers locally while keeping page-level citations for citation checks
- Search HR policy documents for specific rules on a machine that never sends files out
- Analyze confidential financial reports offline in an air-gapped environment
- Review due diligence documents across a deal room without cloud exposure
- Onboard new people onto a project by letting them ask questions of the vault's documents and code
Models Under the Hood
as of 2026-10-16
Limitations
- Airgap is a single-machine, local-first tool: retrieval, generation and citations all run locally, with no hosted account or cloud workspace.
- Local generation depends on your hardware and requires models to be downloaded and cached before offline use works — the first run includes a model download, not just an install.
- Vaults are ordinary files in the app's data directory and are protected at rest only by whatever disk encryption your machine already runs (FileVault, BitLocker or LUKS), so an unencrypted disk means unencrypted vaults.
- A connected-model mode exists but is off by default, and enabling it sends questions off the machine to a hosted API.
as of 2026-10-09
Verification history
We have re-verified Airgap 5 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-checked, vendor evidence unchanged
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-checked, vendor evidence unchanged
Free to cite with attribution — this page re-verifies continuously.
Where the pricing makes sense
The company stage and team size where Airgap's pricing actually pencils out — and where peers do it cheaper.
Airgap's pricing fits teams whose volume aligns with the published tiers. Compare against the alternatives listed below for stage-specific value.
Setup time & first value
How long it actually takes to get something useful out of Airgap — broken out by persona, not the marketing-page minute.
Fastest path is downloading the desktop app for macOS, Windows or Linux and adding your first vault — that gets you to a working index in minutes, but offline use only begins once the local models are downloaded and cached, which adds a one-time download on your connection. Expect the first hour to cover install plus model download; after that, adding documents to a vault and asking questions is
Switching to or from Airgap
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From cloud document chat tools: export the files you already own as PDF, Markdown or text and add them to an Airgap vault, which rebuilds a private vector index locally.
- →From a folder of unmanaged research files: point Airgap at a project's PDFs, Markdown and notes and let one vault replace the ad-hoc folder plus manual keyword search.
- →From an on-prem vector database setup: use Airgap's vault as the container for both documents and their private index, skipping the separate embedding and retrieval plumbing.
- ↗To a hosted document AI workspace: export the source documents out of your vaults and upload them to the hosted service, accepting that answers and retrieval then run in the cloud.
- ↗To a shared team knowledge base: move the underlying PDFs, Markdown and text files into the team platform, since Airgap vaults are single-machine rather than a multi-user workspace.
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Airgap”, and we withheld 6: 6 could not be judged, because “Airgap” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Airgap.
Official links
Tools that pair well with Airgap
Common stack mates teams adopt alongside Airgap, with the specific reason each pairing earns its keep.
Humata AI
Humata AI is a PDF AI that answers questions across your documents with cited links into the exact source section — citations you can trace
Wisedocs
Wisedocs turns messy claim files into medical chronologies, summaries, and defensible reports for insurance, legal, and IME reviewers.
Casper AI
Chrome extension that summarizes web pages, PDFs, and news into short, shareable notes
Featured Head-to-Head Comparisons
Airgap vs Genspark
If you handle sensitive documents that must never leave your machine, Airgap is the only choice—its local, offline processing is unmatched. For everyone else, Genspark's freemium model and broad feature set (Sparkpages, AI Slides, no-code automation) make it the more versatile and cost-effective option, especially with recent open-source initiatives.
Airgap vs Mostly Ai
Pick Mostly AI if you need to generate synthetic versions of large datasets for ML or analytics, especially in cloud ecosystems like Databricks or AWS. Choose Airgap if your top priority is keeping confidential documents entirely on-device for chat and summarization—no cloud involvement. They solve different problems; decide based on whether your data is tabular and shareable (Mostly AI) or document-based and strictly private (Airgap).
Airgap vs Push Security
Choose Push Security if your pain is browser-borne account takeover (AiTM, ClickFix, session hijack) or shadow AI — it's a security platform with automated hunting and deep enterprise integrations. Choose Airgap if your problem is confidentiality for document Q&A: it's a local-only tool that never uploads files, perfect for legal/HR/finance. They don't compete directly; your need dictates the pick.
Alternatives to Airgap
View all
Humata AI
Humata AI is a PDF AI that answers questions across your documents with cited links into the exact source section — citations you can trace
Frequently Asked Questions
Used Airgap? Help shape our editorial sentiment research.