Most writing tools that add AI give you one interface. A chat window, a floating assistant, a command palette. You use it or you do not. There is no choice in how you interact with the agent.
Ritemark works differently. You get two distinct interfaces for working with AI agents. The AI Sidebar is a visual chat panel that sits next to your document. The built-in terminal is a full VS Code terminal where you can run any CLI agent directly. Both operate on the same plain Markdown files on your hard drive. You pick the one that fits how you think, or use both at the same time.
The AI Sidebar
The sidebar opens on the right side of your workspace. It is a visual conversation interface where you type a request and the agent responds. You can ask it to rephrase a paragraph, insert a new section, restructure a long document, or answer a question about your content.
At the top of the sidebar sits the agent switcher dropdown. Three agents are available: Claude Code, OpenAI Codex, and OpenCode, which runs on your own API keys from Anthropic, OpenAI, Google or OpenRouter. You can switch agents in the middle of a conversation, and each answer shows which agent wrote it.
Plan mode is one of the features that makes the sidebar feel safe for big changes. Switch the mode picker to Plan only, and the agent lays out its intended steps before touching anything. Plan only works with Claude Code and Codex, and the default mode is Auto, so you turn it on when you want it. You review the plan, approve it, or tell the agent to adjust. This matters when you are asking for significant restructuring of a long document. You see the intentions before they become edits.
When the agent needs clarification, it asks. Interactive questions appear mid-conversation with selectable options. Pick the answer that fits, or type a custom response. The agent adjusts its approach based on what you tell it instead of guessing and getting it wrong.
A context usage bar at the top of the sidebar shows how much of the agent's memory window you have used. It turns yellow at 60%, red at 80%. When the bar fills up, you can use /compact to summarize the conversation and free up space, or start a fresh thread.
The Built-in Terminal
Below or beside your document, you can open a terminal panel. This is not a simplified shell or a plugin. It is the actual VS Code terminal running inside Ritemark. Every command you would run in a standard terminal works here.
Install a command-line agent and type its name: claude starts a Claude Code session, gemini starts Gemini CLI, codex starts OpenAI Codex. The agent sees your entire project folder, reads any file, and makes changes where you point it. All of this happens in real time on the files sitting on your drive. No upload step, no sync, no copy-paste.
This is exactly how developers have been working with AI agents for months. VS Code and Cursor gave programmers purpose-built environments where agents operate on real files through a real terminal. Ritemark brings that same architecture to everyone who writes. The terminal is there for people who want it, and it is the real thing.
Because the terminal is standard, you are never locked into a specific tool. When a new CLI agent ships next month, install it and run it. Done. Ritemark does not need to add support for it. The terminal already supports everything.
Multiple Agents, No Lock-in
Ritemark does not bundle a house AI. It does not bet on one provider and force you to use whatever model they chose for you. In the sidebar, you switch agents with a dropdown. In the terminal, you install and run whatever you want.
You can even use multiple agents on the same project at the same time. Ask Codex to clean up a paragraph in the sidebar while Claude Code restructures your outline in the terminal. They both read the same files. No configuration required, no conflicts. Each agent works on the filesystem directly.
This is a direct consequence of how Ritemark is built. Your files are plain Markdown on your drive. Both interfaces access them through the filesystem. The sidebar runs Claude Code, Codex and OpenCode, and any other command-line agent that can read files works in the terminal once you install it, including the ones that ship tomorrow.
Your Files Stay Local
Your documents are plain .md files in a folder you choose. Ritemark has no cloud for your documents. There is no account to create, no subscription to cancel, no cloud service that could disappear and take your documents with it. Your files sit in Finder or Explorer, and they will be there in ten years regardless of what happens to Ritemark.
Local storage is also what makes both AI interfaces work so well. Whether you use the sidebar or the terminal, the agent accesses your files through the filesystem directly. It does not download them from a server. It does not sync them from a cloud. The file on your drive is the file the agent reads. That directness keeps everything fast and predictable.
One thing to be clear about: local-first applies to Ritemark's storage. Whatever the agent reads from your folder is sent to the AI provider you chose (Anthropic, OpenAI, or the provider behind your OpenCode key), under your account and its terms. Ritemark does not proxy those requests, and your conversations are saved on your computer, not on a Ritemark server. Anonymous usage analytics are on by default and can be switched off in Settings.
Why This Matters for Writers
Developers got agentic AI workflows early. Tools like VS Code with Claude Code, Cursor with its built-in agent, Windsurf with its own take on the concept. Writers got... chatbots in browser tabs. Copy the text in, paste the result out, manually merge the changes. That workflow breaks as soon as your project is more than a single document.
Ritemark gives writers the same agent-powered architecture that developers have had. Your files are real files. The agent reads them directly. Changes happen in place. And if you never want to open a terminal, the sidebar gives you the full agent experience through a visual conversation.
The two interfaces exist because people think differently. Some prefer clicking through a visual chat. Others prefer typing commands. Some use the sidebar for quick edits and the terminal for deep structural work. Ritemark does not force a preference. Both paths lead to the same result: your agent working on your files, with you in control.
FAQ
What is the AI Sidebar in Ritemark?
The AI Sidebar is a visual chat interface on the right side of the workspace. You type a request and the agent responds. It supports agent switching, plan mode, interactive questions, and context management.
Can I use different AI agents in Ritemark?
Yes. The sidebar has a built-in agent switcher with Claude Code, OpenAI Codex, and OpenCode (with your own API keys). The terminal supports any CLI agent you install, including Gemini CLI.
What is plan mode?
Plan mode lets you see what the agent intends to do before it makes changes. Switch the mode picker to Plan only (Claude Code and Codex) and the agent shows its steps for approval before acting. Auto is the default mode.
Is the terminal a real terminal or a simplified shell?
It is the real VS Code terminal. Every CLI command works, including Claude Code, Gemini CLI, Codex, or any other tool, once you install it.
Do I need to use the terminal?
No. The AI Sidebar provides the full agent experience through a visual interface. The terminal is there for users who prefer command-line workflows.
Are my files stored in the cloud?
No. All files are plain Markdown on your local hard drive. Ritemark has no cloud for your documents. What the agent reads is sent to the AI provider you chose, but Ritemark does not proxy those requests, and conversations are saved on your computer. Anonymous usage analytics are on by default and can be switched off in Settings.
Is Ritemark free?
Yes. Ritemark is free and open source. The AI is not: Claude Code needs a paid Claude plan or an Anthropic API key, Codex a paid ChatGPT plan, and OpenCode your own API keys.