When you use the AI sidebar, you are interacting with artificial intelligence. Ritemark shows the selected runtime and model in the composer and provides an AI information button for details.
AI output can be inaccurate, incomplete, or misleading. Review factual claims, citations, calculations, commands, and file changes before you rely on, publish, or act on them.
If an agent gives you output that is inappropriate or harmful, you can report it to the Ritemark team. See Report AI Output.
Which AI Service Is Used?
Ritemark does not send AI requests through a Productory server. It launches an AI runtime on your computer, and that runtime connects to the provider using the account or API key configured for it.
| Runtime in Ritemark | Provider or service | Authentication used by the runtime |
|---|---|---|
| Claude Code | Anthropic | Claude account sign-in or an Anthropic API key |
| Codex | OpenAI | ChatGPT sign-in or OpenAI API authentication |
| OpenCode | The service selected in Ritemark: Google, OpenAI, Anthropic, or OpenRouter | Your API key for the selected service |
The exact model is the one shown in Ritemark's model picker. As of v1.8.6 that selection is pinned on every path an agent can run: chat turns, queued prompts, comment tasks, flows, and scheduled agents. If the runtime resolves a different model than the one you picked, the conversation states so rather than substituting quietly, and the provenance line under each response names the runtime and model that actually produced it. Available model names can change as the installed runtime and Ritemark's shared model catalogue are updated.
What May Leave Your Device?
Your Markdown files remain normal files on your computer. Using cloud-connected AI is a separate data flow: content needed for a turn can be sent to the selected runtime and provider.
Depending on the task and the context visible in the composer, this can include:
-
Your prompt, plus hidden instructions Ritemark adds so the runtime can complete the task
-
The active file, including its path and content the agent reads
-
Selected text, its file path, and a short surrounding context window
-
Files or images you attach when the selected runtime supports them
-
Shared browser context, such as the page URL, page summary, and an optional screenshot for supported runtimes
-
Tool results and recent conversation context, including a handoff when you switch runtimes in the same chat
An agent may also read other files in the open folder or use tools and commands while completing a task. What it can do depends on the autonomy you selected (Manual or Auto) and the approvals you give. While a conversation is in plan mode, the runtime blocks file writes outright, and approving a plan hands over at the autonomy you already chose.
Removing a context chip excludes that item from the turn where the control is supported. It does not prevent an agent from later reading the same file in your folder with an allowed tool.
What Stays Local?
Ritemark stores and edits your files locally. Productory does not receive a copy of your folder through an AI proxy, and Ritemark does not provide cloud storage for your documents.
This does not mean all AI content stays on the device. A cloud-connected runtime must send prompts and relevant context to its provider to generate a response. The provider handles that data under the terms and privacy settings attached to your account.
You can use Ritemark as a Markdown editor without using the AI sidebar.
Publishing to Google Docs is a separate choice, outside the AI features. When you choose Create or Sync, Ritemark sends that document to your own Google Drive. See Publish to Google Docs.
Ritemark Analytics
Ritemark analytics is separate from AI-provider processing. When analytics is enabled, the desktop app sends anonymous product-usage events to PostHog's EU service. The setting is enabled by default and can be turned off in Ritemark settings.
The product events cover app sessions, feature use, selected agent use, and reactions. They do not include your prompts or file contents. If you explicitly submit written feedback, the feedback text you enter is sent with that feedback event.
Ritemark stores a random anonymous identifier on your device so sessions can be counted without a user account.
Review and Approval
Permission and approval controls help reduce unintended file changes or commands. They do not verify that an AI answer is true, lawful, unbiased, complete, or suitable for your situation.
Before using AI output:
-
Check important facts and sources against reliable originals.
-
Inspect file changes and commands before accepting them.
-
Do not send confidential or regulated information unless your organisation permits the selected provider and account configuration.
-
Apply meaningful human review before publishing or making consequential decisions.
Report AI Output
As of v1.11.0, you can report AI output that is inappropriate or harmful to the Ritemark team. There are two ways in, and both open the same Report AI output window:
-
Report AI issue in the status bar at the bottom of the window, beside the AI Ready indicator. It is always visible there, whether or not the AI sidebar is open. Clicking it opens the AI sidebar with the report window.
-
AI information (the ⓘ button beside the composer) has a Report AI output section with a link of the same name.
The window shows exactly what the report will contain. Under Included with your report you see the two lines Ritemark adds: the Ritemark version with your platform (for example Ritemark 1.11.0 (darwin) on a Mac, or (win32) on Windows), and the time you opened the window, in UTC. Under What went wrong? there is an empty box where you paste the AI output and say what is wrong with it. It holds up to 8,000 characters.
Nothing else goes into the report. Ritemark takes nothing from your conversations, your documents, or your account. Nothing is filtered out either, because the report window never has access to them in the first place. The window starts empty every time you open it.
Click Open email to hand the report to your email app as a new message addressed to info@productory.eu, the address the Ritemark team monitors. You can read and change the whole message there, including the two lines Ritemark added. Nothing goes out until you press Send in your email app. Ritemark only knows that it passed the report to your email app, not what happened next, so it never tells you the report went out. The window says Your email app is opening and reminds you that the report has not gone out yet.
Your report stays in the window whatever happens, and Copy report under the text box copies all of it. You need it in two cases:
| What the window says | What to do |
|---|---|
| No email app is set up on this computer | Copy the report and email it to info@productory.eu yourself. This is a normal outcome, not an error. |
| This report is too long to open in an email app | Email links can only carry a limited amount of text, so Ritemark won't open your email app and risk cutting the report short. Copy it and email it yourself. |
The address is always on screen under the text box as Write to us: info@productory.eu. If no email appears, see Report AI issue doesn't open an email.
Note: In v1.11.0, Report AI issue opened the AI sidebar but not the report window while the sidebar was showing first-run onboarding or agent setup. This is fixed in v1.12.0, so the report window opens before you have set up an agent too.
Provider Information
Provider data retention and model-training choices can depend on your account type and settings. Check the current provider documentation for the account you use.