FAQ

First-week questions. Linux-first, local hatch, bring your own keys.

Canonical URL: /docs/help/faq

About the product

What is Cadence?

A personal AI assistant that runs on your computer. You hatch it, name it, and it keeps identity and memory on disk. It can take actions through skills and tools, with your approval when those actions matter. Overview: What is Cadence.

How is this different from a chat website?

A chat website answers in the tab. Cadence has tools (it can act on your machine and on services you connect), memory that persists across conversations, and a name and behavioral rules you can edit. 0.1 lives on Linux desktop plus the cadence CLI. Channels (Telegram, Slack, mail, phone) attach to that same local assistant. See What is Cadence.

Is Cadence free?

The Linux app is yours to install. There is no Cadence Cloud signup and no Cadence-billed model meter in 0.1. You pay whatever your chosen model provider charges (or nothing extra if you run a local Hive).

Why does it use tokens when I'm not chatting?

Background work (heartbeat, memory compaction, titles) can call the model you configured even when the chat box is idle. That usage hits your provider, not a Cadence invoice. You can ask the assistant to reduce heartbeat frequency, skip compaction, or point those jobs at a cheaper profile.

What platforms does it support?

Linux is first (AppImage and apt). Reach the assistant from the desktop app and the cadence CLI. Windows, macOS, browser hatch, and mobile are later.

Where should I host my assistant?

On your Linux machine. That is the 0.1 default. Workspace and credentials stay on disk. The assistant is available while that computer is awake. There is no public Cadence Cloud signup.

Self-hosting options (local vs advanced) are sketched under Hosting options. Start local.

Can I use it on my phone?

Not as a native 0.1 app. Phone and other channels can attach later to the assistant you already hatched. See Channels.

About privacy and data

Is my data safe?

On 0.1, workspace, memories, and credentials live on your machine. Conversations still go to the model provider you configured so it can generate replies. That is the trade-off. Details: Privacy and data.

Does Cadence train on my data?

No. Cadence does not use your chats to train or fine-tune a Cadence model. Your model provider has its own terms. Read those directly.

Does Cadence collect telemetry?

Only if you opt in. Settings → Privacy has optional usage analytics (token counts and feature adoption, no message content) and crash diagnostics. Both default off.

Can my employer see what I do?

Cadence does not ship an employer admin dashboard for your local hatch. If you run it on a work computer or network, that environment can still see outbound traffic to your model provider. Use your judgment.

What happens if I delete the app?

The instance directory under ~/.local/share/cadence/assistants/ stays until you remove it. cadence retire archives a workspace before removal. Electron userData lives under ~/.config/@aevumcadence/linux. See The workspace and Security best practices.

Can other people access my assistant?

Only if you connect a channel and grant access. Trusted contacts can be limited. Unverified senders get a restricted path. See The permissions model.

About capabilities

What can it do?

Cadence is a local assistant plus a skill catalog, not a fixed feature checklist on this page. For what ships:

• Skills and tools (how capabilities attach)
• Features (integrations, skill authoring, browser, attachments)
• Skills reference (per-skill pages)

On a running assistant, ask it to list installed skills. That live list beats any remembered catalog.

Can it access my files?

Files inside the workspace are in-scope. Files on the rest of the host need an explicit approval (read vs write, once vs a rule). See The permissions model.

Can it send email as me?

If you connect a mail skill and approve the send. Draft-then-approve is the usual path. Connecting a provider is a credential flow, not a secret pasted into chat. See Skills reference for the mail skills that have pages.

Can I teach it new things?

Yes. Tell it facts in conversation (memory). Edit workspace files (SOUL.md, USER.md). Build a custom skill. See Skills and tools and Your first skill.

Can I use it offline?

The workspace is local. Thinking still needs a model. A local Hive (OpenAI-compatible on your LAN or this machine) can reply without a public internet path. A hosted OpenAI-compatible provider needs a network path to that API.

About the assistant itself

Can I change its name?

Yes. Say "I want to rename you to Bob," or edit IDENTITY.md in the workspace.

Can I change its personality?

Yes. Tell it what you want, or edit SOUL.md. Changes take effect on the next conversation. See The workspace.

What AI model does it use?

The one you configured. Bring your own OpenAI-compatible endpoint or a local Hive. There is no Cadence-billed default cloud model in 0.1.

Does it have feelings?

No. It can sound like it does. It is a language model with a workspace costume.