Repository navigation
FAQ
No. Use the Claude API, Gemini or an OpenAI-compatible service with an API key instead, or Apple Intelligence on this Mac (see Settings and Configuration). Without any AI provider, @open, @note, @reminder, @calculate, @agent (needs a coding agent such as Claude Code), @doctor's checks, voice input and pinyin still work; @improve, @translate and @question need one, and @config needs Apple Intelligence or a provider.
On a Mac with Apple Intelligence (Apple silicon, macOS 26 or later), yes: choose "Apple Intelligence (on this Mac)" under Settings → AI Provider. It's free and nothing leaves the Mac, but the model is smaller and less accurate than the cloud models, and doesn't speak every language (not Hindi). See Settings and Configuration → Apple Intelligence on this Mac.
Yes, that's how it starts: a new install has English only, and letters are typed directly. Chinese (pinyin), Korean and Hindi are language packs you add under Settings → Input → + if you want them; French needs no pack. Under "Translate & Rewrite", pick the languages you translate into.
The app is free. Your AI provider bills you by token, to your own account; Apple Intelligence on this Mac costs nothing. Requests are made only when you press the action key; running the same sentence again uses the cache; the more rewrite styles, the longer the output.
Yes: while the results of @improve or a prompt command are up, type what to change ("shorter", "more formal", "in Japanese") and press ⇧⏎; the new results take their place. ⏎ inserts the selected result instead, and Esc goes back to the ones before. @translate has no refine: type to edit the sentence, and ⏎ translates it again. See Commands → Refining the results.
From 0.6.4, ⇧⏎ sends what you typed after AI results (an instruction, or the next question after @question), and ⏎ always inserts the selected result. In 0.6.3, ⏎ sent it.
Yes: once @question has answered, type the next question and press ⇧⏎. The panel becomes a conversation of up to 5 rounds, and each round's answer can be inserted with its digit. See Commands → Conversations.
Yes: after the last round, or with ⌘⏎ at any time, ⌘1, ⌘2 … carry it on in a coding agent installed on this Mac, Claude or ChatGPT on the web, Gemini, or copy it as Markdown. Settings → General → "Continue a conversation in" picks which ones and their order. See Commands → Continuing elsewhere.
Yes: describe it, as in @config candidates in a column or @config Chinese in WeChat. ⏎ shows each change, old → new; ⏎ again applies them, and ⌘Z (or @config undo within 10 minutes) puts them back. Keys, custom commands, extensions and plugins are set only in the settings window. See Commands → @config.
Type @doctor ⏎: it checks the setup on this Mac, lists the problems first, each with a short fix, and fixes the safe ones with ⏎. See Commands → @doctor.
Yes: select it, type @improve, @translate or @question and press ⏎; the result you pick replaces the selection. See Commands → On the selected text.
Type @question with nothing after it: your last 5 questions (or conversations) are listed, and picking one shows its saved answer without going online (⌘R asks again; it was R up to 0.6.3). Settings → General → "Remember recent questions" turns this off.
Yes: click a row of the candidate panel, as you'd press its digit. The row under the pointer is highlighted. The panel never takes the focus, so you keep typing in the same field.
From 0.6.3 that's the default (Horizontal), like the system's own input methods: ← → choose, ↑↓ or - = turn the page. Choose Vertical under Settings → Input → "Candidate layout" for a column.
In an @ command, ⏎ runs the command, and ⇧⏎ inserts what you wrote as typed. Without a command, ⏎ is a plain Return. See Commands.
No. If what follows @ isn't a command name (@john), it is inserted as typed.
No. From 0.6.4, @ typed right after a letter, a digit, ., _ or - is just text: bob@example.com, ssh user@host and git@github.com:you/extensions.git type as usual, in every app, terminals included. See Commands → The command list.
After the command is picked, press ⇧⏎: the draft is inserted as typed, @agent included. You can also type punctuation right after @agent. (@claude is no longer a command from 0.6.4, so it's typed as is.)
Yes: Settings → Agent → "@agent hands it to": Claude Code (the default), Gemini CLI, Kiro CLI, or a command of your own, which gets what you typed as its last argument. The page shows whether each is installed. Only Claude Code can work in the background (@tasks); the others open in Terminal. See Commands → @agent and @tasks.
Yes: Settings → Apps → "Add App…", then "No AI in this app" (or the input menu's "This App: " → No AI). AI commands there send nothing and say so; typing and the other commands work. Apps can also have their own starting language, rewrite styles and typing out. See Settings and Configuration → Per-app settings.
Turn on Settings → AI Provider → Privacy → "Hide sensitive information before sending": they're replaced with placeholders such as [EMAIL_1] on the Mac and put back in the results. It can't recognize names; add them under "Also hide these words". See Privacy.
That's intentional: during secure input, keys go in as typed and nothing is sent to the AI. See Privacy.
Run the installer again ("Install AllInOneIME" from the disk image): it adds AllInOneIME back to your input sources. See Installation and Troubleshooting.
macOS doesn't show third-party input methods' options in System Settings. Open "AllInOneIME Settings" (Spotlight finds it), type @settings, or use the input method's icon in the menu bar → "Settings…".
No, the next request uses the new settings.
Yes, from 0.6.4: a role (role_arn with source_profile), SSO and credential_process profiles work as well as static keys. Credentials are fetched when a request needs them, kept in memory only and renewed before they expire, so there are no temporary keys to copy into ~/.aws/credentials. When an SSO sign-in ends, run the aws sso login --profile X command it gives. Profiles with mfa_serial aren't supported. See Settings and Configuration → AWS profiles.
The API key row in Settings → AI Provider says so: "✓ Saved · ends in a1b2" (the key's last 4 characters), "✓ From $OPENAI_API_KEY" for a key in your shell, or "No API key yet". A key saved before 0.6.4 shows just "✓ Saved". Remove… asks before it removes a key. See Settings and Configuration → API keys.
Yes: put the bundle in a Git repository and add it under Settings → Custom Commands → Extension Sources → "Add from Link…". It clones with your own git and SSH setup, shows what the bundle runs before anything loads, and checks once a day for new commits; nothing changes until you click Update. See Commands → Extension sources.
Settings → AI Provider → Model Settings → turn off "Send temperature", or put "temperature": null in config.json.
Rewrites that only change punctuation or repeat another line aren't shown.
With maxTokens left at its default, the limit is raised automatically for long input; if you set it yourself, increase "Up to … output tokens" under AI Provider → Model Settings.
When nothing is being typed, letters are inserted directly.
Yes. Terminals handle ⌘V themselves; to hand the clipboard's text to a command, press ⌃V in the command. Answers with several lines go in as a paste, so the shell doesn't run each line. With Terminal's "Secure Keyboard Entry" on, nothing is sent to the AI.
Check the notice in the candidate panel: most likely microphone access, or the speech model isn't downloaded yet. See Voice Input.
macOS remembers the added input method, microphone access and preferences by bundle ID; changing it would mean adding and authorizing it again.
Yes: turn on "Type results out" in Settings → General, then "Make occasional typos and fix them". Now and then a key next to the right one on your keyboard layout is hit and noticed up to three letters later, a letter or a space is typed twice, or a short word is typed, deleted and typed again; each is deleted and typed right. A few in 100 characters, never at the start of a line. In code, web and email addresses and numbers, a typo is fixed at once and is only ever a wrong letter or digit; Chinese, Japanese and Korean only hesitate; never in terminals.
In a field that takes typos back (Apple's text views such as TextEdit and Notes, browsers, most editors), typed to the end, or with another key to type the rest at once, the text is exactly the result; Esc (or leaving the field) in the middle of a typo takes the typo out first.
Typos are only made where the input method can take them back. Each time, the first typo is a plain one (a wrong letter corrected in one go) until the field has shown it takes a typo back. A field that ignores where to replace text keeps that first typo (the text ends up one character off), and its app is remembered and gets no typos again. Settings → General lists those apps under the typos switch; to let them try again, click "Forget apps where typos couldn't be fixed", or in Terminal:
defaults delete com.aipinyin.inputmethod.AIPinyin typeOutNoTyposAppsFrom 0.6.3 on: Settings → About → "Check for Updates", then "Update to "; it also looks once a day by itself and tells you, never installing until you click. Older versions: install the new disk image over it. See Installation and Troubleshooting → Updating.
From 0.6.3 on, AllInOneIME is signed with a Developer ID, differently from earlier versions, so after updating from 0.6.2 or earlier (to 0.6.3 or 0.6.4), macOS asks once for each key saved in the keychain ("AllInOneIME wants to use your confidential information…"). Enter your login password and click "Always Allow". Updating from 0.6.3 to 0.6.4 doesn't ask.
Type @reportissue and a short title, ⏎ to see what's added, ⏎ again to open the GitHub page, where you submit it; or use "Report an Issue…" in the input menu or Settings → About. See Commands → @reportissue.
Open "Install AllInOneIME" and click Uninstall…, then delete the data folders; see "Uninstall" in Installation and Troubleshooting.