100% client-side parsing
Capture and semantic analysis run inside your browser. Your full conversation never has to leave the page to be understood.
Engine runs in the extension, not on our servers
ATLAS is built for power users who refuse to hand their full AI transcripts to yet another cloud. Capture and parsing run locally. Only structured excerpts reach your private dashboard.
Capture and semantic analysis run inside your browser. Your full conversation never has to leave the page to be understood.
Engine runs in the extension, not on our servers
We do not upload or persist your full chat logs. Only structured concepts, tasks, and metadata reach your dashboard. Text excerpts are truncated (about 600 characters max per anchor).
Full transcripts stay on your machine
The extension only requests storage and tabs, with host access limited to supported AI platforms and your ATLAS dashboard. No broad « read all websites » access.
Manifest V3 · Explicit AI host list
For developers and security reviewers who want the details.
All communication between the extension, dashboard, and our API uses HTTPS (TLS). Industry-standard transport encryption end to end.
After local parsing, the extension sends session metadata, semantic anchors, action vectors, and short summaries to your account via an authenticated API key. Summaries may be generated server-side via Groq using only the structured context sent, not your full raw transcript.
We use Supabase (hosting & database), Stripe (billing), and Groq (summaries & embeddings). We do not sell your conversations. See our privacy policy for the full list and purposes.
Read privacy policyYou can request access, correction, or deletion of your data at any time (GDPR and similar rights). Contact us and we process verified deletion requests. No hidden copies sold to data brokers.
Common patterns in AI tooling vs. our approach.
We do not store your full raw transcripts. We store structured data you chose to capture: concepts, tasks, metadata, and truncated excerpts needed for search and your wiki.
Session metadata (platform, URL, date, message count), semantic anchors, action vectors, and generated summaries. Anchor content is capped at roughly 600 characters. Your full conversation is never sent as a single blob.
Summaries and embeddings are generated via Groq API calls using only the structured context we send. We design the pipeline to minimize payload size. Check Groq's terms for their retention policy; we do not use your chats to train our own models.
Only you, via your authenticated account. Access controls and row-level isolation apply on the backend. Our team accesses data only when required for support or security, with least privilege.
Yes. Remove the extension from chrome://extensions. Your dashboard account remains unless you ask us to delete it. You can export or delete data by contacting us.
Install the extension, capture locally, and send only what you need to your dashboard.