Files stay on your Mac
Your library, vectors and originals live in the app’s data folder. The public gateway stores no corpus and no vectors.
Local-first · macOS 14+ · MCP connector
Drop in documents, PDFs, images, video and audio. Ask in your own words. Memverix indexes everything on your Mac and takes you to the exact timestamp or page, so your assistant can cite the real thing.
Development test build · ad-hoc signed, not yet notarized · details
Ask in your own words
You rarely remember file names. You remember “the part about cricket”. Pick a kind of evidence and watch Memverix go straight to it.
Illustrative demo with sample data. Results anchor to real timestamps and pages in the app.
Drop anything
Drag in what you have. Text is extracted, scans are read, speech is transcribed and video is sampled, all on your Mac, all into one searchable index.
How it works
Choose files or drag them in. Memverix keeps managed copies and fixes each page and timestamp anchor so citations never drift.
A local worker extracts text, runs OCR, transcribes speech and embeds the result. Your library and vectors never leave the machine.
Search yourself, or let an authorized assistant call the connector. It fetches the specific passage, frame or clip, not your whole library.
Connect to ChatGPT
Memverix speaks MCP, the open Model Context Protocol. Connect it to ChatGPT as a custom connector and your assistant can search your library and pull back the exact evidence, with the page or timestamp attached.
https://memverix.evognix.com/mcpCustom connector setup, not an app-directory listing. Memverix has not been approved by or listed in any ChatGPT directory.
When did the founder talk about cricket?
memverix.search“founder talks about cricket”3 resultsmemverix.fetchfounder-interview.mp4 · 26:04evidenceAt 26:04 in founder-interview.mp4, he says cricket is how he learned to ship fast.
Private by design
Your library, vectors and originals live in the app’s data folder. The public gateway stores no corpus and no vectors.
Embeddings, OCR and transcription run locally. Memverix never calls a frontier-model API to index your files.
An assistant gets the passages, frames and clips it asks for, with provenance, not a copy of your library.
Approve each connection on your device and revoke it any time. Deleting a source never deletes your original file.
Read the Privacy Policy and Terms of Service, including how we handle Google sign-in data.
Supported formats
Text extraction with page anchors; scanned pages read with OCR where available.
Visual and text search. The original image comes back as evidence.
Speech transcribed on your Mac and linked to timestamps.
Indexed through sampled frames and speech. Not every frame is analysed; warnings show per source.
Choose the build for your Mac. We’ll highlight the one we think fits; change it if we’re wrong.
Not sure? Apple menu → About This Mac. “Chip: Apple M…” means Apple Silicon.
.sha256 file. Check with shasum -a 256 -c. Release pageThis build is ad-hoc signed and not yet notarized by Apple. macOS may say it can’t verify the app. Unzip and drag Memverix to Applications. If macOS blocks it, open System Settings → Privacy & Security, scroll to Security and choose Open Anyway (on older macOS, right-click → Open works). We’ll move to notarization and say so here when it lands.
FAQ
No. Import, extraction, OCR, transcription and embeddings run on your Mac. When you connect an assistant, only the evidence it requests (a passage, a frame, a clip) is returned to it through the authenticated connector.
The current build is ad-hoc signed but not notarized by Apple. Open System Settings → Privacy & Security, choose Open Anyway next to Memverix, then confirm. Or run xattr -dr com.apple.quarantine /Applications/Memverix.app in Terminal. This is a one-time step per download.
No. Indexing and search never call a frontier-model API. Your assistant, such as ChatGPT, uses its own account when it reasons over the evidence.
No. Video is indexed through sampled frames plus transcribed speech. Memverix shows extraction warnings per source so you know what was and wasn’t covered.
FFmpeg, Poppler and Tesseract are used for media extraction, PDF page rendering and OCR, and are not bundled yet. You can install them with Homebrew. The first launch also downloads the local runtime and model weights.
No. You add Memverix as a custom connector in developer mode. We make no claim of directory approval.
Delete the source in the app. This removes its managed index data but never your original file. Revoke any assistant connection from the app at any time.
Put your files in once. Ask in plain words. Jump to the exact moment.
Download for Mac