What's in the box
Chauncy is a focused application — semantic retrieval over the documents you choose to keep, with optional AI on your own terms. This page lays out what's actually in the Windows installer.
Search
Chauncy reads what you ingest, breaks it into passages, and indexes each passage by meaning using a local embedding model. When you search, Chauncy returns the documents whose content most closely matches your query — whether or not your words appear in them.
Plain-English queries match by meaning. A search for "natural camouflage" can find an article about zebra stripes that never uses the word "camouflage." A search for "advice about handling difficult conversations" can find an email about an argument with a colleague that shares no keywords with your query.
Type a query that starts with # or ~ followed by a term, and Chauncy switches to a fast string scan instead of semantic search. Use it when you know the exact word or tag you're looking for.
A search result is a source document, ranked by its best matching passage. The excerpt shown on the result card is the strongest match within that document. Clicking opens the full stored view so you can read the surrounding context.
Your notes and reactions to a document become part of the searchable collection. A query can surface a document because your annotation on it matched, not just the source text.
Ingestion
Chauncy ingests material you point it at. It does not scan your hard drive, monitor your activity, or pull from the cloud without your action.
Add individual files or whole folder trees. Supported formats: PDF, Word (.docx), PowerPoint (.pptx), Markdown, HTML/HTM, plain text, and email files (.eml, .mbox). Recursive ingestion through subfolders is supported.
Paste a URL. Chauncy fetches the page and indexes its content. Useful for capturing articles, documentation, blog posts, and reference material you want to keep accessible.
Direct IMAP is best suited to a dedicated reference mailbox, not your live inbox. You decide what gets indexed by choosing what lands in that mailbox.
Configure a folder for periodic ingestion. Chauncy scans it roughly every two minutes, ingests new supported files, and moves processed files to an archive folder you specify. This is an intake queue, not two-way file sync.
Dictate searches and notes directly using your machine's speech recognition. Voice input requires a working WebView2 with speech recognition support and may require network availability depending on your system. English (US) is supported.
Chauncy indexes a local copy of the text it extracts. Your original files stay where they are, except for files moved through a watched-folder workflow. Searches continue to work even if the original source file is moved or deleted.
Annotations
Every document you ingest can carry your annotations — freeform notes attached to the source. Disagreements, confirmations, questions, conclusions, cross-references, anything worth saying.
When you add or change an annotation, Chauncy reprocesses the parent document and includes the annotation text in the search index. A query for an idea you wrote in a margin note can surface the document the annotation lives on.
Annotations travel with the document they're attached to. Open the document later and the annotations are there. Search for content related to your annotation and the document comes back with the annotation visible.
Optional AI
Chauncy's search runs locally and needs no AI provider. When you want a language model to summarize results, draft from sources, or answer questions across your documents, turn AI synthesis on and connect a provider. Until then, nothing leaves your machine.
Bring your own API key and pay Anthropic directly.
Bring your own API key and pay OpenAI directly.
Run a model on your own hardware. No API key, no recurring cost, nothing leaving your machine.
Point Chauncy at any local or remote endpoint that speaks the OpenAI API.
For cloud providers, only the excerpts Chauncy selects to answer your specific question are sent to the provider, and only when you ask. Your full document collection, index, and annotations stay local. For Ollama and custom local endpoints, nothing leaves the machine at all.
You pay the provider directly for what you use. Chauncy doesn't mark up provider costs, doesn't route your requests through our servers, and doesn't see your prompts or responses.
Storage
Chauncy is built around a local database. Your documents, search index, embeddings, and annotations all live on your machine, in a directory you can locate, back up, and inspect.
Chauncy doesn't require an account, doesn't phone home, and doesn't track usage. Buy the installer, run it, use the software.
The embedding model that powers semantic search runs on your machine. It ships with the installer; no network call is required to index your documents or run a search.
Chauncy includes a backup utility in Settings. Create a compressed backup of your full database, list and manage past backups, or restore from a backup to roll back the state. Backups are .tar.gz archives stored where you choose.
Chauncy stores a copy of the text it extracts from each source. If you delete or move the original file, search still works against Chauncy's stored copy. Chauncy does not currently warn you if a source file goes missing on disk; the document remains visible in the index until you remove it through Chauncy.
By design
Chauncy is small on purpose. The capabilities below are not in the software — not as oversights, but as choices.
Chauncy doesn't browse the web on its own, operate accounts, send email, buy things, delete your files, or chain background actions.
Chauncy is not a backup tool, not a file synchronizer, and not a replacement for your file system. Your files live where you put them. Chauncy indexes what you ingest.
Chauncy doesn't ask you to tag, backlink, or organize. The connections it surfaces come from the meaning of what you've written, not from a graph you maintain.
Chauncy supports one collection per installation. Many ingestion sources, one searchable library. Multiple libraries or projects are not in the launch build.
Requirements
Chauncy ships as a Windows desktop application. The installer handles required components where they're not already present.
Updates are released as new installers. Auto-update is not in the current build — when a new version is available, download and install from the Chauncy website.
macOS support is in development. A waitlist is open in the site footer.
On the roadmap
Chauncy is actively developed by a small team. A few capabilities not in the current build that we're working on:
We don't pre-announce dates, but these are the priorities the team is working through.