Everything Valence can do.

Chatting

The main screen is a conversation. Type in the box at the bottom and your AI replies above it.

By default, press Enter to send (you can switch this to Ctrl + Enter in settings if you'd rather use Enter for new lines). Press Esc to stop a reply that's still being written.

Watching it write

Replies arrive complete and cleanly formatted rather than flickering as they generate. If you prefer to watch the text stream in live, there's a one-click toggle on the reply.

Reasoning

Some models show their working. When they do, Valence tucks that Chain of Thought into a small collapsible note so it never buries the actual answer.

Saving & finding chats

Nothing is saved to a server. Conversations live as files on your computer.

Editing & branching

You can change what you asked and explore a different answer without losing the original.

Hover a message you sent and choose to edit it. Saving re-runs the conversation from that point, creating a branch — you can move between branches to compare where each one led. Ctrl + Shift + R regenerates the most recent reply.

Files, images & voice

You can give Valence documents and pictures, or talk instead of typing.

Attachments

The paperclip attaches text files, PDFs, and images. Long documents are split into chunks automatically so they fit the model's context. Images need a model that can see — most current cloud models can, and Valence tells you when the selected one can't.

Voice

Ctrl + M toggles voice input. Ctrl + Alt + T is push-to-talk and works even when Valence isn't the active window.

The built-in AI

Valence ships with an AI that runs on your own machine. No account, no key, no internet needed.

It's set up during first run and is the right choice for most people — it's free, private, and works offline. Gemma 4 is the recommended everyday model and the only family suitable for Kids Mode. Qwen 3 is stronger at code, maths, and non-English conversation, but is more verbose and isn't offered for Kids Mode.

The model library shows how much memory each model needs and dims any that your machine can't comfortably run, so you can't accidentally choose one that will crawl.

Cloud providers

Valence can also talk to the major cloud AIs using your own key.

Supported: OpenAI, Anthropic (Claude), Google (Gemini), xAI (Grok), OpenRouter, Ollama, plus Azure OpenAI and Vertex AI for business setups. You pay the provider directly; Valence takes no cut and never sees your conversations.

Keys are encrypted on your machine using Windows' own protection and are never uploaded. Switch provider or model from the dropdowns at the top of the chat.

Step-by-step help getting a key lives in the setup guides.

The status badge

A small pill beside the wordmark always says which AI you're using and where it runs.

Click it for the full picture, including how to switch.

Profiles

A profile is a saved way of working — a name, an icon, a personality, and a preferred AI.

Create one from the profile menu with + New. Give it a name, pick an icon, and optionally write System Instructions telling the AI how to behave (up to 10,000 characters). Switch profiles from the same menu or with Ctrl + P.

Keeping profiles separate

Each profile can isolate its memories, so a work profile never recalls anything personal. Profiles can also be exported to a file and imported on another machine.

For a child, use the Kids Mode wizard instead — see Kids Mode. A child's profile is created there, not here.

Appearance

Themes, text size, motion, and how the window behaves.

The Theme pane is grouped into Style, Motion, Text & Input, and Window. You can change the colour theme, adjust font size and family, reduce animation, and control whether Valence floats above other windows.

Ctrl + Alt + O shows or hides the window from anywhere, even when you're in another app.

Tabs

Run several conversations side by side.

Local memory

With memory on, Valence carries what matters from one conversation to the next — all stored on your machine.

Memory browser

Everything Valence remembers, in one place — readable, searchable, deletable.

Each memory shows its title, a summary, when it was saved, and its topics. Filter by text or date, browse by topic, or delete any memory permanently. Nothing is hidden from you.

Importing chats

Bring your history from ChatGPT, Claude, or Gemini.

Choose the provider, sign in through the browser window that opens, and confirm. Your conversations become searchable memories on your machine — they are not uploaded anywhere.

ChatGPT also accepts a ZIP export if you'd rather not sign in through Valence. Gemini has no bulk export, so it's read one conversation at a time and takes longer.

Knowledge folders

Point Valence at folders of your own documents so it can answer from them.

Add a folder as a knowledge source and its contents become searchable context for your questions. Sources can be attached per profile, so a work profile can read work documents without touching anything else.

Kids Mode

A protected space for a child, created through a guided six-step wizard.

Start from the profile menu → + New → Use the Kids Mode wizard. The wizard walks you through consent, setting a parental password, adding each child (name, age, icon), previewing the safety behaviour, choosing controls, and creating the profiles.

What gets locked

While a kids profile is active, Valence hides settings, profile creation and editing, saving and exporting conversations, the memory browser, usage, and model comparison. Switching profiles requires the parental password. A Kids Mode badge shows in the title bar.

Which AI a child can use

Only the built-in local model or Vertex AI. With the local model, a child's conversations never leave the computer.

Child controls

Each child gets age-appropriate defaults that you can tune.

If a child is in distress

If a conversation contains self-harm or crisis language, Valence surfaces the 988 Suicide & Crisis Lifeline and flags the conversation for you in the Parent Dashboard rather than trying to handle it alone.

Parent Dashboard

A calm overview of each child's activity, with anything worth a look floated to the top.

Open it from the menu and enter your parental password. It fills the window, with a rail down the left:

The dashboard locks itself after a few minutes of inactivity, and re-locks if the window is minimised while it's open. Everything stays on your PC.

Where your data lives

On your computer, as ordinary files. There is no Valence account and no Valence server.

Kids conversations can additionally be encrypted at rest, so nobody else using the same computer can read them. Settings → Storage shows where everything lives and lets you back it up or wipe it.

Local-only mode

Switches off every optional online feature so nothing can leave your machine.

You can turn it on during first-run setup or later in settings. Note that it does not block the things you explicitly ask for — downloading a model, or importing your history from another service, still reach the internet by design.

Licence & trial

Free for 7 days, then a one-time purchase. No subscription.

The trial is the full app — nothing is held back and no card is needed. After it ends, enter your licence key to keep going. Every feature is included in the one price; there are no tiers and nothing is gated.

Turning on Advanced

Valence hides its power tools until you ask for them.

At the top of the settings sidebar there's an Advanced switch. Turn it on to reveal Triggers, Abilities, MCP, Automations, Diagnostics, Contexts, and other expert panes. Leave it off and settings stays short and calm — which is where most people should leave it.

Abilities Advanced

Small built-in actions the AI can run mid-conversation.

You don't need to type these — the AI uses them when they're useful. You can switch individual abilities off.

MCP tools Advanced

Connect outside tools that follow the Model Context Protocol.

Add a server and its tools become available to the AI as abilities. Anything a tool wants to do asks for your confirmation first. Tools never connect on their own — you turn them on deliberately.

Comparing models

Ask the same question of several AIs at once and read the answers side by side.

Pick the models you want, send one prompt, and compare. Useful when deciding whether a cloud model is actually worth it over your local one.

Keyboard shortcuts

Logs & diagnostics

If we ever need to help you troubleshoot, this is where the details are.

The Log Files panel shows the folder path (clickable to open it), how long logs are kept, and buttons to copy the path or open today's log. Logs stay on your machine — nothing is sent to us unless you choose to send it.

Troubleshooting

Windows blocked the download or the app

Expected for a new app with no reputation yet. Keep the file from your browser's download menu, then on first open click More info → Run anyway. It only happens once.

The first reply is slow

A local model warms up the first time you use it after opening Valence. The status badge shows what's happening. After that it stays warm.

A cloud provider rejects the key

Re-copy the key — a stray space is the usual cause — and check the account still has credit.

It forgot something

Memory must be switched on, and it saves what looks important rather than everything. Open the Memory browser to see exactly what it kept.

Still stuck

Email help@helixailabs.com.

This handbook describes Valence 1.4.2. Features are listed as they appear in the shipping app; panes marked Advanced only appear when the Advanced switch is on.