Settings

Most of Wanigan needs no choosing, so its settings are few. Open them with On a Mac, ⌘,·; on Windows and Linux, Ctrl+,, the rail’s Settings, Wanigan 2 › Settings…, or Go › Settings; Go › Accounts opens Accounts directly.

In the list, ↑ and ↓ move between sections, and Home and End go to its ends. Every section is in the palette too (On a Mac, ⌘K·; on Windows and Linux, Ctrl+K): type its name, or “Live view settings” and “Voice input settings”.

More on tours in Tours and How this works.

Both are kept by the app, not the core, in electron/settings.json in Wanigan’s data folder. How notifications arrive is on The board.

Folded under them, Hosted environments says how a site’s Dev, Test and Live are opened: read-only, and only when you do. These are kept in electron/settings.json, like General. Turning the view on is in Turning it on.

Voice input turns speech into text on this computer; the audio never leaves it. This section picks the speech model and shows what your operating system says of the microphone.

The choices are kept in voice/settings.json, and the models in voice/models/, in Wanigan’s data folder. More in Speaking instead of typing.

The daily choice is kept in electron/settings.json, and the last check in electron/updates.json. Installing stays a download you choose: see Updating.

Every Claude Code and Codex account Wanigan found, each with who it is signed in as according to the agent, its limits and the projects that use it. Add account beside an agent makes and signs in a new one; Sign in, Make default, and More › Rename or Remove from Wanigan act on one row; Check again asks every agent once more; How this works draws the rules.

Each sign-in stays in the account’s own folder, where the agent keeps it; Wanigan never holds, reads or copies a password or a token. Its database keeps the list of accounts and each project’s choice. Everything else is in Accounts and models.

Every skill your agents can see, grouped by where it lives: personal, in a project, from a plugin, or synced from claude.ai. Reading is free; nothing is copied or removed until you click, after seeing what changes.

Skills stay in the agents’ own folders; Wanigan keeps none of its own. Built-in skills and Codex’s plugin skills are not listed.

Every MCP server each account and open project has, read from the agents’ own files, and a store of twelve well-known ones. A change runs the agent’s own command, as the account you chose, only after you have seen it exactly.

Servers are kept in each agent’s own settings. Values that look like keys or tokens are never shown, and Wanigan never asks for or keeps a key. Gemini CLI’s servers are copied into Wanigan’s Gemini home only when they hold nothing that could be a credential, and this section says so.

Models that run on this computer. Under LM Studio, Its server says whether it runs, with Start its server. Models Wanigan knows lists each with the agent it runs with and whether that pairing is proven: Get shows the size and asks first, Stop stops a download, and Carry on resumes one. Other models on this computer lists what LM Studio, Ollama or NVIDIA PAIR has besides.

Models are kept in LM Studio’s own folder, ~/.lmstudio/models, not in Wanigan’s; Wanigan keeps 10 GB of your disk free, and a model is removed in LM Studio itself. The whole story is in Models on this computer.

Phone access and each paired phone are kept in the core’s database, a phone’s key only as a hash; a new phone can act until you turn that off. This Mac’s push key is in phone-push.json, readable only by you. Setting it up is on The board, and what a phone may reach in Security and privacy.

Under Connection, paste a TypeSafe API key and click Save and test, or export TYPESAFE_API_KEY in your shell; Test the connection and Forget the key follow. So far counts calls, failures and the estimated cost, and In each project sets each project to Off, Reads or Reads and accepts. Once a key is set, Jev’s health shows beside the title.

A saved key is jev.key in Wanigan’s data folder, readable only by you, and a key in your shell wins over it. Each project’s mode is in the core’s database, and starts on Reads. All of it is in Jev.

What one project needs differently, such as a work account or its own branches, is in its own settings. Click the project’s name in the project bar and choose Project settings….

The switch and the pickers apply at once; the name, key and setup command wait for Save. All of it is kept in the core’s database.

• Jev’s mode is on the Jev strip above the board, and in Settings › Jev.
• Pause is in the project bar (Pause and Decisions).
• The live view’s site, its helper and its hosted environments are under Site settings… in the live view’s site menu (Turning it on).
• Before and after screenshots are one switch for every project, in Settings › Live view.
• Decisions, the rules told to every session, are a view of their own (G then D).