Getting started

Beta

Download Wrengle, open a folder as your vault, and write your first note — that's the whole setup. Everything else on this page is worth exploring once you're comfortable.

Download Wrengle

Open wrengle.com/download to get the current build. The page only offers the exact-named desktop installers uploaded to the currently configured GitHub release—normally latest, or an exact incident-recovery tag during a temporary pin—and marks a platform unavailable when its installer is missing. Linux isn't a v1 target, so it won't appear there.

Open or create a vault

A vault is the local folder Wrengle uses for your notes and its own app-managed metadata. Choose one you can back up.

Write your first note

Use the file tree or the command palette to create a note. Wrengle stores it as a local file, with its own app metadata kept beside the vault.

What happens the first time you open Wrengle

Before the normal workspace appears, Wrengle runs a short access check for whatever your configured features need. On macOS, approve microphone access when asked. System audio stays off until you explicitly run its short audible test under Settings → Privacy & Data → Desktop access. Wrengle also loads your saved provider, account, integration, action, plugin, and workflow secrets from the operating-system keychain at the same time.

Each macOS keychain prompt is scoped to a single read. Allow covers just that one read; Always Allow remembers it for the current app identity. This first pass only makes those secrets available to the app for the session ahead. It does not contact a provider, refresh OAuth, run an action, start an agent, or begin recording. Those all still wait for you to ask for them, later.

On Windows, this check works differently. The current unpackaged Windows build reports native capture access as Not required. It does this without probing a device or checking the desktop-wide microphone switch, and capture features never show a Windows permission dialog. If Windows or your chosen device later blocks capture, that device-open attempt fails without any prompt. Open the Windows microphone Settings link, always available under Settings → Privacy & Data → Desktop access, to fix it.

If a supported startup access check is denied, Wrengle opens in Limited mode instead of interrupting you again later. Local vault and note work remains available. Only the specific blocked capability is unavailable. The startup panel and Settings → Privacy & Data → Desktop access show what's blocked and offer System Settings, Restart Wrengle, or a direct retry where that helps. See What you need for the complete platform-by-platform detail.

Sign-in is optional and isn't part of first-run setup. Beta accounts are available through email or Google. The optional status-bar account control opens the system browser; credentials stay in the operating-system keychain, and your vault stays local either way. See Accounts and sign-in.

Find your way around

The workspace has three core regions, plus an optional fourth:

the project/file tree;
the editor area;
the agent panel;
a far-right utility panel for History, Search, Backlinks, and contributed tools, when one is open.

The agent and utility panels are independent. You can keep both open, and closing either one leaves the other in place.

If you need support, open Settings → About, choose Copy diagnostics, and include that privacy-safe summary in your message. It contains the Wrengle version, platform, and local background-storage status, but excludes vault paths, note and transcript content, accounts, secrets, and raw logs.

Try Quick Dictation

Open Settings → Dictation → Microphone and voice setup, choose Dictation microphone, and use the five-second Test microphone action. It stays local and never starts transcription. Then focus a note, hold Cmd+Shift+D on macOS or Ctrl+Shift+D on Windows, and release when you finish speaking. The voice panel should name that note as its destination and commit your partial result on release.

Dictation speech and voice analysis default to Auto, but they resolve independently. Speech can use a connected speech-provider key confirmed available in your OS keychain, otherwise it uses Local Whisper. Analysis uses the most recently connected OpenAI or Anthropic AI key, then checks OpenAI and Anthropic in that order, with the provider's reviewed default model. Verbatim Quick Dictation bypasses analysis. Without a usable AI key, analysis pauses for setup and sends nothing. Key validity is checked only when a request runs, so an invalid, revoked, or quota-limited key can still fail later.

Set up generative AI

Generative AI needs a route before it will run anything. On a clean install, assistant chat starts on Wrengle AI · Fast and shows its destination, storage, and charging disclosure before the first managed request. Upgrading does not replace an existing explicit provider or Auto choice.

Under Settings → Models → Provider, keep or choose Wrengle AI to use models Wrengle runs and pay with prepaid account credits, or choose OpenAI or Anthropic to use a key of your own. A key stays in the operating-system keychain and its requests never pass through Wrengle. Wrengle AI needs no key; its prompts and replies transit Wrengle and OpenAI for generation but are not stored in the account ledger or application logs. Auto still considers only configured provider keys and never selects Wrengle AI.

For assistant chat, you can make the same choice from Run with at the bottom of the Agent panel. The picker remains there during setup: OpenAI and Anthropic accept a missing key inline, Wrengle AI offers account sign-in, and external coding agents show their own authentication methods above it.

Signing in does not include free credit. Credits are bought on your account at wrengle.com and spent as you make requests. Nothing renews on its own and there is no overage plan: when the balance runs out, requests stop until you top up. Settings shows what is left. Fast, Balanced, and Deep are text-only managed tiers, and none falls back to a key of your own. Wrengle AI and credits publishes their exact mappings, rates, and current deployment-availability rules; AI and privacy sets out exactly what each route sends and to whom.

Configure the language, Verbatim or Light edit cleanup, and response speed under Dictation. Local speech keeps microphone audio on the device; cloud speech sends it to the effective provider. See Dictation and voice control before you use voice commands or decide which data boundary you want.

Try a workflow

Open Workflows from any New tab (+) menu or the command palette. Start with a blank workflow or a starter, then use Dry run to inspect it or Run to execute an autonomous-safe draft immediately. To run it on a daily or specific-days schedule, choose Automate…. Wrengle pins the current draft as that automation's live version. Read Workflows before you rely on scheduled behavior.

Read limitations before relying on a workflow

Wrengle is in beta. Review Known limitations before relying on dictation, voice commands, workflow automation, transcription, AI, plugins, or download/access behavior.