Index files on your Mac. Videos, audio, PDFs, slides, screenshots, scans — understood by multimodal AI and searchable by meaning, in plain English (even if the file is in Japanese). Your originals stay on disk; AI traffic uses your own Gemini API key, billed by Google directly.
Free to start. No credit card required.
You've recorded years of meetings, podcasts, lectures, demos, decisions. Saved decks, scans, screenshots, contracts. They sit as opaque blobs that search can't see into. The next time you need that one moment — you can't find it.
All that knowledge is locked inside files that search can't reach.
Other tools convert your video to text, then search the text. GoldenRetriever uses multimodal AI to understand the original media directly — the visual context of a presentation, the tone of a voice, the diagram on a whiteboard.
The AI processes visual content natively. Slides shown on screen, product demos, whiteboard sessions — it understands what's being shown, not just what's being said.
Podcasts, coaching sessions, meeting recordings, voice memos. The full audio signal is embedded — emphasis, context, and meaning that a transcript would lose.
PDFs, Word documents, PowerPoints, images, and plain text. Every format your team uses, indexed and searchable alongside your media.
The result is a complete knowledge layer across everything your organisation produces. Ask a question and get an answer that draws on a training video from January, a slide deck from March, and a voice memo from last week — with sources and timestamps for each.
Real searches from real people. If you have files, you have a use case.
“Find every wedding kiss shot from the last three years.”
Wedding videographer
“Find the shot of the red car at golden hour.”
Filmmaker / DIT
“Find that screenshot of vertical tab navigation I saved in spring.”
Designer
“Which interview mentioned pricing pushback?”
UX researcher
“What did we decide about the auth migration in the eng sync three weeks ago?”
Engineering lead
“Find the French Airbnb screenshot, travel docs, foreign tax forms, expat paperwork, German rental contracts.”
Expat / digital nomad
“I know I wrote this somewhere.”
Writer
“Find the clause about indemnification across 200 PDFs.”
Lawyer
“Find the moment in the deposition where they said X.”
Litigator
“40 hours of recordings from one investigation — find every mention of the shell company.”
Investigative journalist
“Which paper said the effect size was 0.3?”
Academic
“Find the photo of the cracked beam from the Oak St job.”
Construction PM
“Did I keep the receipt for the dishwasher?”
Anyone, honestly
One worked example, in depth — from pitch day to client disputes to qual research to building something your competitors can't replicate.
The Pain
Senior planners and account directors lose 8–12 hours per pitch assembling case studies, results decks, and credentials material. Junior staff can't help because the relevant work is buried across hundreds of decks, recordings, QBRs and emails — and they don't know what's there.
Ask GoldenRetriever
Show me every campaign result for a B2B SaaS client in the last 24 months — the challenge, our solution, the measured outcome, and which slides we used.
What You Get Back
Sourced answers from QBRs, results presentations, strategy decks and post-campaign retros — with timestamps and slide numbers. A draft credentials section, written by the agency's own work, in under an hour.
The Pain
Critical client decisions happen in moments — during a Zoom call, mid-screen-share — and become contested weeks later. 'Did they actually approve that budget?' turns into account-level risk and uncomfortable email chains.
Ask GoldenRetriever
Find the call where the CMO said yes to the £80k extension. Show me the moment, with their face when they said it.
What You Get Back
The exact recording, the exact timestamp, with full multimodal context — words, tone, expression. Account managers get certainty. Disputes get settled in seconds. The agency's recall of its own commitments becomes provable.
The Pain
Qualitative analysis is the most expensive single activity in agency life. A 90-minute focus group typically takes 6–10 hours of senior time to synthesise properly. And the most valuable insight isn't in the transcript — it's in everything the transcript loses: the hesitation before the price answer, the expression when the alternative concept appeared, the body language of the dominant voice in the room, the moment the energy in the group shifted.
Ask GoldenRetriever
Show me every moment in our concept-test focus groups where participants visibly hesitated before answering questions about price.
What You Get Back
GoldenRetriever watches the video, hears the tone, and indexes both verbal and non-verbal context. Hours of senior analysis collapse to seconds. The roughly 70% of communication that has always lived only in the analyst's memory becomes searchable across every focus group, every depth interview, every vox-pop the agency has ever recorded.
Why This Defends Your Agency
For agencies running qual research for clients, this turns commissioned research from a cost line into a defensible IP layer that compounds over time. For research agencies, it is the foundation of a premium analysis service no transcript-based competitor can match. Either way, the multimodal layer is the moat.
The Pain
Clients are using ChatGPT to write their own brand copy. AI tools are eating into traditional agency margins. The agency narrative is being reframed as a workflow that can be brought in-house.
Ask GoldenRetriever
Show me every campaign, message, asset and visual identity decision we've made for this client in the last five years — and recommend a creative direction for the next quarter that's consistent with all of it.
What You Get Back
A defensible, monetisable AI service that clients literally cannot replicate in-house. They don't have the archive. You do. The agency stops being a workflow that's vulnerable to AI commoditisation, and starts being the AI platform.
Why This Defends Your Agency
This is the punchline. Multimodal embedding of your accumulated work becomes a productised service — 'your brand's AI, built on five years of our work for you' — that's structurally impossible for a freelancer plus ChatGPT to replicate.
GoldenRetriever turns hours of unwatchable, unsearchable content into answers you can find in seconds.
Scrubbing through hours of rushes and archived footage to find the right shot or soundbite.
Search your entire archive by meaning. 'Find the interview where she talks about sustainability' — instant results with timecodes.
New starters can't access institutional knowledge locked in hundreds of training videos nobody rewatches.
New hires ask questions in plain English and get sourced answers drawn from your complete training library.
Years of client videos, campaign decks, and strategy presentations scattered across drives. Impossible to search.
Instantly find when the CEO discussed brand positioning, or pull insights from across dozens of campaign reviews.
Hundreds of recorded sessions with clients. Valuable insights buried in audio files nobody has time to relisten to.
Ask 'What did we discuss about hiring strategy with Acme Corp?' and get the answer with the exact session and timestamp.
Town halls, all-hands, board recordings, onboarding materials — all recorded, almost never rewatched.
Your entire organisational memory becomes searchable. No more 'I think someone mentioned that in a meeting once.'
If you have more than a few hours of recorded media, you have knowledge you can't access.
GoldenRetriever makes it all findable. One app, every format, every answer.
Tell GoldenRetriever which directories to watch. It scans everything — videos, audio, documents, presentations — and starts indexing automatically.
Each file is processed with state-of-the-art multimodal AI. Videos are understood visually, audio is heard natively, documents are read in full. Everything is stored in a local knowledge base on your machine.
Search across your entire library or ask questions in natural language. Get precise answers with source attribution — including timestamps for video and audio — and one-click navigation to the original file.
Ask questions in plain English across your entire media collection. Get sourced answers with exact timestamps for video and audio.
Find content by what it's about, not just what it says. Semantic search understands intent, synonyms, and context across every file type.
AI processes video visually and audio natively — not just transcripts. The diagram on a whiteboard, the tone of a voice, the slide on screen.
Point GoldenRetriever at folders on your Mac and it picks up new files as they appear. Run an explicit reset on edited or removed files (better edit/delete handling on the roadmap).
Your originals stay on disk. To run AI features, the relevant content is sent from your Mac to your chosen AI provider on your own API key — your bill, your terms. The Company never sees your files.
Pick your preferred model for Q&A — Gemini, OpenAI, Anthropic, Ollama, or any OpenAI-compatible endpoint. Your infrastructure, your choice.
Whether you're organising a personal collection or managing terabytes of corporate media, GoldenRetriever grows with you.
Tested with multi-terabyte media libraries
Per-seat Enterprise pricing today; multi-seat admin and SSO on the roadmap.
Billing and account data on Hetzner Nuremberg. SSO and at-rest encryption on the roadmap for Business and Enterprise.
You paste a Google Gemini API key during onboarding. Embedding and AI traffic go directly from your Mac to Google's Gemini API on that key — we never see it and we don't touch the bill.
Powered by Qdrant running locally. Your embeddings and search index live on your machine — not in someone else's cloud. Fast, private, and fully under your control.
Gemini Embedding 2 with 3072 dimensions. True multimodal processing means the AI understands the original media directly — the visual context of a video or the nuance of a voice recording isn't lost in transcription.
GoldenRetriever is in public beta. Subscribe during the beta and your price never rises — we may raise prices for new customers after general availability. All plans include full search + AI Q&A. You only pay for GoldenRetriever; your GCP costs are billed directly by Google.
Experience the full multimodal magic
Everything a Mac-based business needs — founding-member rate
For teams that need more — founding-member rate
Talk to us. Roadmap items below are on the path, not all shipping today — scope and timing are negotiated per contract.
Free to start. macOS 14 or later. No credit card required.
Windows and Linux on the roadmap.