Choose a passphrase to encrypt your conversations. Astrata cannot read your chats — only you can, with this passphrase. Use anything memorable — a word, a phrase, a number you won't forget.
AI with organization
🔒 Your chats are end-to-end encrypted.
Astrata cannot read them — only you can.
By signing in you agree to our and
To sign in with a different account, you'll need to sign out of Google first.
This opens in a new tab — come back here when you're done.
Sign out of Google ↗These are issues we know about and are actively working on. Submit new bugs via .
On some devices, the mobile view shows "already in sync" but doesn't reflect the latest desktop changes. Workaround: tap the sync icon to force a pull, or use Settings → Reset cloud data to resync cleanly.
In rare cases, refreshing the page very quickly after a new response before it finishes saving to cloud can result in the response not appearing after reload. Workaround: wait for the sync icon to show green before refreshing.
The "This Month" usage in Settings now syncs across devices via your account. Previously it was per-device only — if you see $0 on a device that was recently signed in, try sending one message to trigger a sync.
Version history snapshots are now taken at most once per hour. Content created within the same hour window won't have a separate snapshot — but the main cloud save always reflects the latest state.
Astrata is a personal AI research workspace. This policy explains what data we collect and how we use it.
All conversation data is encrypted in your browser using AES-GCM 256-bit encryption before it is stored on our servers. The encryption key is derived from a passphrase that only you know — it is never transmitted to Astrata, never stored on our servers, and never accessible to us in any form. This means Astrata's operators physically cannot read your conversations, even with full access to the server infrastructure. The only way to read your data is to know your passphrase.
If you are using Astrata without your own API key, your messages are processed through our servers using a shared trial key. During this trial period, conversations are not end-to-end encrypted — they pass through Astrata's infrastructure to reach the AI provider. We do not log or read conversation content, but we cannot offer the same technical guarantee of privacy as the fully encrypted experience. Trial access is limited to Claude Sonnet 4.6 and a set number of messages. To unlock full end-to-end encryption, unrestricted model choice, and unlimited usage, add your own API key in Settings.
If you use your own API key (OpenAI or Google Gemini), your conversation content is sent directly from your browser to the AI provider — it does not pass through Astrata's servers at all. For Anthropic (Claude), a server proxy is required due to browser security restrictions, but no content is logged.
To improve Astrata's layout and functionality, we use session recording tools (such as ContentSquare or similar services). These tools record interactions such as clicks, scrolls, and navigation patterns to help us understand how the interface is used. All recorded session data is fully obfuscated — no conversation content, personal information, or typed text is captured. Only anonymized UI interaction data is collected for product improvement purposes.
Encrypted workspace data is stored on Netlify's infrastructure. In addition to your current workspace state, up to 5 rolling encrypted snapshots are retained to support version history and data recovery. Shared thread content (created via Share Thread) is stored unencrypted on the same infrastructure for the duration the share is active. No data is sold or shared with third parties.
To request deletion of your account and all associated data, email pmaass@gmail.com. Requests will be fulfilled within 30 days.
This policy may be updated. Continued use of Astrata constitutes acceptance of any changes.
Last updated: February 24, 2026
Astrata is a personal AI workspace provided as a beta service. By accessing or using Astrata, you agree to these Terms of Use. If you do not agree, do not use the service.
These Terms govern your access to and use of Astrata ("the Service"), operated by Paul Maass ("Operator"). The Service is provided on an as-is, beta basis for personal and professional research purposes. These Terms may be updated at any time; continued use constitutes acceptance.
You must be at least 18 years old to use the Service. By using Astrata, you represent that you meet this requirement. The Service is currently available by invitation or open beta at the Operator's discretion.
You agree to use Astrata only for lawful purposes and in accordance with these Terms. You must not:
Astrata uses third-party AI models (including those from Anthropic, OpenAI, and Google) to generate responses. AI-generated content may be inaccurate, incomplete, or misleading. You are solely responsible for verifying any information provided by the AI before relying on it for any purpose. Astrata makes no representations about the accuracy, reliability, or fitness for any particular purpose of AI-generated responses.
You retain ownership of all content you submit to the Service. By using Astrata, you grant the Operator a limited, non-exclusive license to process your content solely for the purpose of providing the Service (including routing messages to AI providers). Conversations processed via the trial API key pass through our infrastructure. Conversations sent using your own API keys go directly to the provider. See our Privacy Policy for full details.
A limited free trial is provided to allow you to evaluate the Service. Trial usage is subject to rate limits, restricted model access, and may be modified or terminated at any time without notice. For continued, unrestricted use, you are responsible for obtaining and managing your own API keys from Anthropic, OpenAI, or Google, subject to each provider's terms and costs. Astrata is not responsible for charges incurred with third-party AI providers.
THE SERVICE IS PROVIDED "AS IS" AND "AS AVAILABLE" WITHOUT WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT. THE OPERATOR DOES NOT WARRANT THAT THE SERVICE WILL BE UNINTERRUPTED, ERROR-FREE, OR SECURE.
TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, THE OPERATOR SHALL NOT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR PUNITIVE DAMAGES, OR ANY LOSS OF DATA, REVENUE, OR PROFITS, ARISING OUT OF OR IN CONNECTION WITH YOUR USE OF OR INABILITY TO USE THE SERVICE. IN NO EVENT SHALL THE OPERATOR'S TOTAL LIABILITY EXCEED THE AMOUNT YOU PAID FOR THE SERVICE IN THE PAST 12 MONTHS (OR $50 IF YOU HAVE NOT PAID ANYTHING).
You agree to indemnify and hold harmless the Operator from any claims, damages, or expenses (including reasonable attorneys' fees) arising from your use of the Service, your violation of these Terms, or your violation of any third-party rights.
The Operator reserves the right to suspend or terminate your access to the Service at any time, for any reason, without prior notice. Provisions that by their nature should survive termination shall survive.
These Terms are governed by the laws of the State of New York, without regard to conflict of law principles. Any disputes shall be resolved in the state or federal courts located in New York.
Questions about these Terms may be directed to pmaass@gmail.com.
Last updated: February 24, 2026
Web Browsing
Voice (ElevenLabs)
Reliability
PDF Export
Local Model Diagnostics
**bold text** for section headings instead of markdown # syntax. These are now recognised as headings (h3) so they appear in the Navigator and as Follow-up jump links.↑219 ↓551 · 23 tok/s) so you can see how fast your hardware is running.Reliability
Image Quality Fix
Polish & Accessibility
New Providers
Reliability
Navigator
Known Issues
UI Improvements
Version History
Bug Fixes — Sync & Version Restore
New: xAI Grok Models
GROK_API_KEY env var is set in Netlify.Admin
Bug Fixes
New Features
Polish
Step-by-step tutorials for local models, remote access, and more
Ollama runs open-source AI models directly on your computer. No API costs — your conversations are processed by a model running on your own hardware, not sent to a cloud AI provider.
Go to ollama.com and download the installer for your OS (Mac, Windows, or Linux). Run the installer — it sets up Ollama as a background service.
Open Terminal (Mac/Linux) or Command Prompt (Windows) and run:
This downloads Llama 3.2 (3B, ~2 GB). Other good options:
Run ollama list to see all your downloaded models.
Astrata runs in your browser. By default, Ollama only accepts requests from localhost — not from web pages. You need to set an environment variable to allow it:
Mac / Linux:
Windows (Command Prompt):
Windows (PowerShell):
Open ⚙ Settings in Astrata and scroll to the Local section:
http://localhost:11434ollama list (e.g. llama3.2)Hit Save, then select Local from the provider dropdown. That's it.
Want to use your phone or a second laptop on the same WiFi? You need Ollama to listen on the network (not just localhost):
Then find your computer's local IP address:
192.168.1.50)ipconfig in Command Prompt → IPv4 AddressOn your phone/tablet, open Astrata Settings and set the Local URL to http://192.168.1.50:11434 (use your actual IP).
~60 seconds • highlights every key feature
Core Features
Setting Up Providers
gsk_Option A: Ollama
OLLAMA_ORIGINS=* ollama serve (the OLLAMA_ORIGINS flag allows browser access)ollama pull llama3http://localhost:11434 and model name to llama3Option B: LM Studio
http://localhost:1234 and model name to match what you loadedFree Trial & Privacy
API Keys
sk-ant-sk- — you can only view it onceAIzaAI & Memory
Keyboard Shortcuts
You have 30 free messages to explore before you'll need to add your own API key. It takes about 30 seconds when you're ready.
Your conversations are end-to-end encrypted before leaving this browser — Astrata cannot read them. API keys never leave this device.
Applied to every conversation as your personal context.
Free tier available · Get yours at console.groq.com → API Keys
Runs in your browser → localhost. Requires Ollama or LM Studio running locally. Setup guide →
Your key lives only on this device. Back it up to access your data on another device.
Resets each browser session.
Usage syncs across your devices via your account. Each device tracks its own calls and the highest count is shown.
Local snapshots saved automatically after each AI response. Stored in this browser only — use these to recover a thread if cloud sync fails.
Wipes cloud storage and all version snapshots, then re-uploads your current local workspace as a clean start. Useful if sync is stuck or data is mismatched. Does not affect your API keys or local usage stats — those stay in your browser.
Astrata uses AI and can make mistakes. Please double-check responses.