Privacy policy

Exactly what stays on your machine and what leaves it.

Last updated: October 20, 2026

Overview

Sennoric makes the Solan AI models and the software that uses them, from a coding agent on your computer to chat on the web. Some features run on your device; others use hosted services. This policy explains what stays local, what Sennoric stores, and which kinds of providers process data.

What Sennoric stores

Sennoric's hosted backend stores the data it needs to run your account and hosted features:

  • Account: your email address, password hash or linked Google, GitHub or Discord account ID, verification and reset state, suspension state and appeal records, and the IP address from sign-up and from your latest Google, GitHub or Discord sign-in. We use IP addresses to spot duplicate accounts and abuse. When a new account shares an IP address with an existing one, a team member is alerted to review it; that never suspends an account by itself.
  • Profile and settings: the name you want to be called, custom instructions, response language, profile picture, selected model, notification preferences, and when you accepted the Terms and this policy.
  • Chats: signed-in chat history from the website, the iPhone app and the desktop app's cloud chats. That includes your messages and the files and images you attach, the model's replies, its reasoning text and tool results, and your projects, artifacts, scheduled tasks and the files the model creates. Deleted chats stay in the trash for 30 days.
  • Memory: facts saved to your memory, which are added to the model's instructions in your chats. Memory is on by default.
  • Learn AI progress: if you sign in on sennoric.com/learn, the lessons you have finished, your best score on each, your XP and your streak. Without signing in, this stays in your browser only.
  • Shared links: when you share a chat, a copy of it, including its photos, that anyone with the link can open.
  • Connected apps: if you connect GitHub or Notion on the website, an encrypted copy of the access token, kept until you disconnect, so the model can read and write there when you ask.
  • API keys and usage: your Sennoric API keys, stored as one-way hashes so they can't be read back, with their status, request and token counts, usage cost, allowance windows and daily usage totals.
  • Billing: plan and subscription status, credit balance and redeemed credit codes, Stripe customer and subscription IDs, and a Stripe payment-method ID if you turn on auto top-up. Sennoric never stores your card number.
  • Safety log: separately from your chat history, Sennoric logs every exchange with a hosted model, from the apps and from API keys alike. Each entry holds your message and the full context sent with it (including custom instructions, memories and tool results), the reply, any tool calls the model made, your account, the API key used, your IP address and the time. Deleting a chat does not delete its log entries.
    • Every four hours, and whenever our team starts a review by hand, an automated process sends recent exchanges (the conversation so far, the message and the reply) to a third-party AI safety classifier. Each comes back safe, and nothing further happens; as an operational error, meaning it could not be classified, which is recorded as a system fault and never as a violation; or flagged, which puts it in a queue until a member of our team reads it and dismisses or confirms the finding.
    • In a signed-in chat (on the website, in the iPhone or desktop app, or in a scheduled task), the model can privately flag your message if it sees a clear, serious violation, such as sexual content involving a minor, a credible threat to harm someone, malware or hacking, help with weapons or serious crime, targeted hateful abuse, or an attempt to copy the model. Fresco rarely does this on its own, so when Fresco declines a request there, Sennoric's servers also privately ask it the same question. Either way, a flagged exchange goes to the same queue for human review. Flags are never shown in the chat and never suspend an account by themselves.
    • Safe exchanges, operational errors and dismissed findings are deleted after 30 days. Unreviewed exchanges and pending or confirmed findings are deleted after one year; we keep them that long for human review, safety enforcement and legal compliance.
    • A confirmed finding may lead to suspension of the account, which also cancels any subscription. Suspended users can appeal.
  • Teams and preferences: organization and team membership, roles and invitations, and your email and announcement preferences.
  • CLI sign-in codes: short-lived codes for signing in the CLI, deleted once they expire.
  • Forms you send us: Team and Enterprise requests from the Teams page (your name, work email, company, role, message and the setup you chose), which are also emailed to our support inbox and kept for one year; and Open Day challenge entries (your name, email and entry).
  • Rate limits: counts of recent requests per account or IP address, used to limit abuse and deleted within a day after their time window ends.
  • App error reports: when the iPhone app hits an error, it sends the app and OS version, your device model and the error details. We keep these for 90 days.
  • Service records: a record of each reply the hosted models generate (the model, timing and any technical error), and any change our team makes to a plan, allowance or credit balance, including who made it and when.

Announcement subscriptions can exist without an account. Deleting your account also ends announcement emails to its address.

What stays on your device

The Sennoric CLI and desktop app work on your computer. These stay there unless noted. The CLI described here is its next release; the CLI you can install today differs in a few ways, listed under The CLI you can install today.

  • Your code: the agent reads, edits and runs files on your computer. File contents leave it only when they're part of a prompt to the model you chose, or when you use a feature that sends them elsewhere, such as web search, cloud artifacts, a vision server, a connected app or phone pairing (see the next section).
  • CLI memory: memory, learned preferences and notes are stored in ~/.sennoric/. Memory in website, iPhone and desktop cloud chats is stored by Sennoric (see above).
  • Keys you add: stored in ~/.sennoric/ in files only your user account can read, obscured but not strongly encrypted, and sent only to the provider or server they belong to.
  • CLI connections: when you connect GitHub, Google, Notion or Slack in the CLI, the token is stored in ~/.sennoric/oauth.json and sent only to that service, never to Sennoric. GitHub, Notion and Slack run as third-party packages that the CLI downloads from npm and runs on your computer; Google uses direct API calls. What each can reach, only when you ask:
    • GitHub: read and write access to your repositories, issues and pull requests (scopes repo read:org read:user).
    • Google: read-only access to Google Drive, Google Calendar events (list, create, delete), and your Google email address and profile. Sennoric does not request Gmail access.
    • Notion: the pages you choose on Notion's approval screen.
    • Slack: whatever the bot token you paste in can reach; it can read channels and post messages.
    Anything they return becomes part of the prompt sent to your model.
  • Local models: point the CLI at a model server on your own computer, such as Ollama, with /endpoint, and inference stays there. Web search, update checks and package downloads still use the network.
  • Browser bridge: the Chrome extension talks to the CLI or desktop app over 127.0.0.1 with a pairing token; that connection doesn't leave your computer.

Features that send data off your device

These send data from your device to Sennoric or to another provider.

Solan models. Your prompt, plus anything included as context (file contents, command output, images or page text), goes through Sennoric's backend at api.sennoric.com to Sennoric's inference servers, which run on rented cloud GPUs. Using a Solan model requires a Sennoric account, signed in or with a Sennoric API key. Chats on the website and in the apps count against your Free, Pro or Team allowance and any credits; API-key requests are paid from credits. In the CLI you can instead send prompts to a server you set up with /endpoint.

Connecting apps on the website, desktop or iPhone. When you connect GitHub, Google or Notion there, the sign-in goes through Sennoric's server, which exchanges the sign-in code for an access token. The desktop and iPhone apps collect the token from Sennoric, which deletes its encrypted copy as soon as they do, or within a few hours if they never do. On the website, Sennoric keeps an encrypted GitHub or Notion token until you disconnect and uses it to read and write there when you ask. These connections ask for the same access as the CLI's, and GitHub also shares your email address.

Signing in with Google, GitHub or Discord shares your email address and account ID with Sennoric.

Voice.

  • Voice input in the iPhone app uses Apple's speech recognition, which runs on your device when it can and on Apple's servers otherwise.
  • Voice input on the website and in the desktop app uses the speech recognition built into your browser or the app's web view, which may send the audio to its maker (Google, for Chrome-based browsers).
  • Read-aloud in the iPhone app and on the website sends the message text through Sennoric to a text-to-speech provider.
  • The CLI doesn't record or transcribe audio.

Web search. When the model searches the web or reads a page for you in a signed-in chat, the search terms or page address go to a web-search provider. In the CLI, search uses Tavily with your own key, or DuckDuckGo without one.

Code sandbox. When the model runs code in a signed-in chat, the code and its files run on a cloud sandbox provider, and the files it produces are stored with the chat.

Phone pairing. When you pair the iPhone app with the desktop app, everything the two exchange, including chats and tool output, passes through Sennoric's servers. It is encrypted in transit and not stored. Sennoric keeps a short-lived record of the pairing.

Screen vision. The CLI has no built-in vision model. Screen and computer-use features send screenshots to the model you choose with /vision.

Chrome extension. The extension asks for access to all sites and for the scripting, tabs, side panel, storage and context-menu permissions, so it can read pages, click, fill in forms and capture the visible tab when you ask. Page content you send goes to the AI provider you set up with your own key; the extension has no Sennoric-hosted model. Your keys, chats and bridge token stay in the extension's local Chrome storage on your computer and are not synced.

Email. Resend sends Sennoric's emails: sign-up verification, password resets, suspension and appeal notices, usage warnings, invitations, scheduled-task results and announcements, plus internal alerts to our team.

Payments. Stripe processes payments for Pro, Team and credit purchases, including auto top-up if you turn it on. Sennoric stores the IDs and status it needs to manage access, never your card number.

Bot checks. Sign-up and sign-in pages use Cloudflare Turnstile, which checks your browser to stop automated sign-ups.

Update checks. The CLI asks the npm registry whether a newer version exists.

Training-data contributions (/contribute). Nothing is sent unless you run /contribute yourself. Before anything is stored, Sennoric's server removes images and files, cuts each message to about 1,000 characters and tool output to 200, and masks text that looks like a key, an email address or a home folder name. The CLI makes the same redacted copy before sending. If the upload fails, that redacted copy is saved to ~/.sennoric/donations/ and is never uploaded later; you can delete it whenever you like. Contributions help train future Solan models.

The CLI you can install today. The CLI on npm and Homebrew today, @axion-labs-ai/quark-cli 2.0.1 (the axion command), works differently from the next release described above. It stores its files in ~/.axion. It can send prompts straight to Anthropic, OpenAI, Google Gemini, Groq, Mistral, OpenRouter, OpenCode, Z.ai or a local Ollama server with your own keys, and speak text aloud through OpenAI. Its Google connection asks for full Drive access, its default screen-vision model no longer exists, and its /contribute sends the full conversation, which Sennoric's server redacts on arrival as described above.

Other companies that process data

  • Cloud GPU hosting (rented servers that run Solan models): your prompts and replies are processed there to generate responses.
  • An AI safety classifier run by a third party: recent exchanges are sent there for automated review, as described above.
  • A web-search provider, a code-sandbox provider and a text-to-speech provider, used by signed-in chats when the model searches the web, runs code or reads a reply aloud.
  • Cloudflare (hosting, storage, network and bot checks on sign-in): cloudflare.com/policies/privacy
  • Stripe (payments): stripe.com/privacy
  • Resend (email): resend.com/legal/privacy-policy
  • GitHub (hosts this website; sign-in and connected apps): GitHub privacy statement
  • Google (sign-in, connected apps, the website's fonts, and our support inbox): policies.google.com/privacy
  • Discord (sign-in): discord.com/privacy
  • Notion and Slack (connected apps): Notion privacy policy, Slack privacy policy
  • Apple (voice input in the iPhone app): Apple's dictation and privacy page
  • Your browser's maker (voice input on the website and in the desktop app), for example Google in Chrome.
  • jsDelivr and unpkg (scripts the website loads in your browser): jsDelivr privacy policy
  • npm (CLI update checks and package downloads): npm privacy policy
  • AI providers and tools you set up yourself, such as a server added with /endpoint, an MCP server, a search key, or a provider key in the Chrome extension or the CLI you can install today (Anthropic, OpenAI, Google Gemini, Groq, Mistral, OpenRouter, OpenCode or Z.ai): your requests go straight to them, so see their own policies.

Your controls

  • Chats: move a chat to Trash from the chat list on the website, desktop or iPhone. Trash keeps it for 30 days; on the website, open Trash and choose Delete forever or Empty Trash to remove it now. A chat in Trash stops its shared links, and deleting it for good removes them. Deleting a chat does not remove its safety-log entries (see above).
  • Shared links: stop sharing any chat or artifact under Settings, Privacy & data, Shared links, or in the iPhone app's Privacy settings.
  • Memory: turn memory off, or edit, delete or clear saved memories, under Settings, Memory.
  • Artifacts, projects and scheduled tasks: delete them in the chat interface. Uploaded and generated files are deleted with their chat.
  • Connected apps: disconnect under Settings, Connections on the website or in the desktop app, under Privacy in the iPhone app, or with /oauth revoke <service> in the CLI. That deletes Sennoric's or your device's copy of the token; to withdraw the app's access completely, also remove Sennoric in that service's own connected-apps settings.
  • API keys: revoke them at sennoric.com/keys. A revoked key stops working immediately.
  • Email: choose which emails you get under Settings, Notifications. Account and security emails always send. If you subscribed to announcements without an account, use the unsubscribe link in any announcement.
  • Download your data: Settings, Privacy & data, Download gives you a copy of your chats, projects, memory, preferences, Learn AI progress and API key names. For anything else we hold, such as safety-log entries, email support@sennoric.com.
  • Delete your account under Settings, Privacy & data on the website, or in the desktop or iPhone app. Sennoric cancels any subscription first, then deletes your chats, files, memory, keys, connections, Learn AI progress and settings, and stops announcement emails to your address. Safety-log entries lose your account, API key and IP address. Routine entries are emptied; pending or confirmed flagged exchanges keep their text, which could still identify you, until review ends or one year passes. Billing records are kept for accounting, unlinked from you, and Stripe keeps its own payment records under its policy.
  • Keep prompts off Sennoric's servers: in the CLI, send them to your own server with /endpoint, for example Ollama on your computer. The Chrome extension only uses providers you add. The desktop and iPhone apps use Solan models.
  • Contributions: nothing is sent unless you run /contribute. Delete ~/.sennoric/donations/ whenever you like.
  • Suspension: if your account is suspended, use the appeal link in the suspension email.

Contact

For questions about this policy or your data, or to ask for a copy or deletion of data not covered above, email support@sennoric.com from the address on your account.