SHIPPED · v1.20.0
Secrets Stay Secret
New
Credentials leave your memories automatically — When a conversation is captured or an AI tool saves a memory, API keys, tokens, passwords, SSH private keys and database passwords move into the Secret Vault, and the memory keeps a {{secret:NAME}} placeholder in their place; turn it off in Settings → Security → Lift secrets from new memories
Recognizes keys from the services you use — OpenAI, Anthropic, ElevenLabs, DeepSeek, OpenRouter, GitHub, GitLab, AWS, Google, Stripe, Slack, Hugging Face, Groq, xAI and more, plus KEY=value lines, --token style flags, Bearer headers and passwords inside database connection URLs; names come from the surrounding text, such as ELEVENLABS_API_KEY
Scan the memories you already have — Scan memories finds the credentials already in your vault, lets you search the results, name each one and choose what to lift, then rewrites the memories in small batches while you keep working
A Secrets view for every key — A new Secrets item in the sidebar lists every stored secret with its kind, provider, how many memories use it and when an AI tool last used it; filter by name or kind, and sort by name, last used or most used
An inbox for anything that needs a decision — Values that need a name, look like a new version of a key you already stored, or might not be secrets wait in an inbox, where you can Name it…, Replace value, Keep both or Restore to memory
Manage each secret — Reveal a value for 10 seconds, copy it (the clipboard clears after a minute), rename it everywhere it's used, replace its value, attach it to more memories, check its access log, or delete it
Add, import and export — Add a secret yourself and attach it to a new or existing memory; import a .env file, pasted text or an encrypted backup with a preview and a choice for names that already exist; export a passphrase-protected backup or a plain .env file
Click a placeholder to copy — {{secret:NAME}} placeholders show as chips in your conversations, and clicking one copies the value
Ask for a key by name — Claude Code, Codex, Cursor and other MCP-connected AI tools get three new tools: vault_secret_list shows the names you have, vault_secret_resolve makes a secret available to the tool's shell, and vault_secret_set stores a key you ask the AI to remember
Values never reach the chat — Instead of the key, the AI tool gets a short-lived file that it loads in the command that needs it, referring to the key only by name, so the value stays out of the conversation, the terminal and the tool's own logs; the file is deleted after 15 minutes
You decide who gets access — By default 1AIVault asks each time, with Allow once, Always allow this secret or Deny; change the default in Settings → Security → When an AI tool asks for a secret, or per secret under AI tool access, and every request is logged
Sealed on this device — Values are encrypted with your system's secure storage (Keychain on Mac), and only the running 1AIVault app can open them
Start from a prompt gallery — A new chat opens with ready-made prompts grouped into Recall, Code, Write, Plan and Learn, such as "What did I decide?" and "Draft in my voice"
Queue or steer while it answers — Messages you send during an answer wait their turn by default; switch to Steer to stop the current answer and send yours right away, or take a queued message back out before it goes
Copy and retry any answer — Hover a message to copy it or answer that turn again
Rename and tidy your chats — The chat list shows a preview of each conversation, chats can be renamed, deleting one asks for a second click, and ⌘N / Ctrl+N starts a new chat
See it working — Answers stream in place with a running "Working…" timer while the AI tool thinks
Improved
Vault exports carry your secrets — Encrypted vault exports include the Secret Vault by default (Include Secret Vault), and importing one adds those secrets to the ones you have; memory packs and passports always leave secrets out
Vault Doctor checks the Secret Vault — It reports whether the Secret Vault is ready and warns when its key can't be opened on this device
Fixed
Fixed Retry in Chat doing nothing — Retry now runs the turn again with the AI tool you have selected
Fixed buttons and dialog titles rendering too large — Small and regular buttons, text fields and dialog titles now use their intended sizes
Fixed AI tools' vault searches losing semantic matches while 1AIVault indexes memories — Searches now get an answer right away instead of waiting behind the indexing
Fixed live import turning itself off after a few errors in a row — It now rides out a busy vault and gets a fresh set of retries after a minute of running cleanly

The Secrets view listing stored keys with their kind, how many memories use each one, when an AI tool last used it and its AI access setting, under an inbox of three secrets that need a decision 
Scan memories for secrets at the Review step, with two ElevenLabs keys found in 12 and 2 memories, both selected with high confidence 
Claude Code asking to use ELEVENLABS_API_KEY, with Deny, Always allow this secret and Allow once, and a note that it gets a file it can load in its shell for 15 minutes 
A new chat opening on the prompt gallery, with Recall, Code, Write, Plan and Learn prompts such as What did I decide? and Draft in my voice