Changelog
Shipped, fixed, and improved
Every 1AIVault release, straight from the release notes that ship with the product.
SHIPPED · v1.12.0
Encrypted cross-device vault sync and task-ready Focus Briefs
New
- Sync Across Devices
- Focus Briefs
- Evidence-Aware Memory
Improved
- Watch scheduled automation live, with status, AI messages, thinking, and tool activity, then open the memories a run classified
- Take control of active schedules — stop, safely stop and delete, or clear failed runs while keeping successful history
- Read more useful Dream diaries, with a clear action, affected memory, and reason for every applied change
Fixed
- Fixed AI tools installed through PowerShell not being found on Windows
- Fixed meaning-based recall failing to start on Windows
- Fixed one unavailable AI tool hiding other installed clients
SHIPPED · v1.11.0
Memory That Thinks — Watch Your AI Recall, and Let Your Vault Dream
New
- Neuron — Watch Your AI Recall: a new screen that records every memory search your connected AI tools run against the vault and lets you replay it stage by stage.
- Semantic recall: searching now finds the right memory even when phrased differently, searches long conversations passage by passage, and pulls in related memories automatically — all on-device.
- Dreaming — overnight vault consolidation: schedule a nightly dream and your AI CLI reviews clusters of related memories, merging duplicates, fixing tags, and linking what belongs together, leaving a dream diary each morning.
SHIPPED · v1.8.1
A focused update that makes links, file paths, and scheduled Claude Code classification safer and more reliable.
New
- Link and File Actions
Fixed
- Fixed **links replacing the vault with a blank or unusable page** — clicking a link in memory details now keeps 1AIVault open and gives you safe actions instead
- Fixed **scheduled classification with Claude Code failing to reach the vault** — unattended schedules can find the required memory tools and save classifications again
New
- Train a Personal Memory Model — fine-tune a compact local model on your memories, entirely on your machine (Apple Silicon Macs), and ask it questions about your own knowledge offline.
- Pick what it learns: your whole vault, selected topics, or a recent time window — with a preview of how many memories are in scope and roughly how many training examples they produce.
- Secrets stay out — API keys, tokens, and private keys are detected and redacted before a single training example is written, with a pre-scan that shows what will be hidden.
Improved
- A brand-new startup experience — the app opens instantly on an animated 1AIVault logo, then fades into a fully-loaded Dashboard with no white flash or half-drawn screens.
- Large vaults stay smooth — heavy background work (statistics, activity heatmaps, memory linking, tier promotion, conversation capture) now runs off the app's hot path.
- Faster Dashboard and sidebar on big vaults — counts and stats load in the background instead of competing with what you're doing.
Fixed
- Fixed the app freezing with a spinning cursor for several seconds after launch on large vaults.
- Fixed a light-theme flash at launch for dark-theme users — the window now opens in your theme from the first frame.
- Fixed the app briefly showing your vault before the unlock screen when Vault Lock is enabled.
New
- Vault Lock — lock 1AIVault behind a passphrase; while locked the app and any MCP-connected AI tool are refused reads (Argon2id, no recovery, idle auto-lock)
- Bring Your Memory — seed the vault from a ChatGPT export, claude.ai memory, an Obsidian vault, or CLAUDE.md/AGENTS.md/.cursorrules, with each entry tagged by origin
- Grok CLI Support — Grok Build joins the supported clients for conversation import and MCP memory sharing via ~/.grok/config.toml
SHIPPED · v1.6.0
Redesigned vault, more MCP clients, better Classify controls
New
- Neural Vault Interface - See the vault through a redesigned Dashboard, global search, activity heatmap, and consistent app surfaces.
- More AI Tool Connections - Install 1AIVault MCP for GitHub Copilot, Roo Code, Qoder, Trae, Factory Droid, Kilo Code, Warp, and Augment, with Cline CLI support across import, chat, and classify.
- Deeper Classify Control - Review forgotten memory, remember or clean up hidden items, rebuild classifications from scratch, and merge topics with selected entries.
Improved
- Choose classification context size for bulk classification and save the preferred default.
- Classification prompts inspect more context when needed and include confidence for every assignment.
- Collect separates browser-extension captures, on-machine imports, and other supported tools.
Fixed
- Cline MCP connection finds current Cline installs and repairs stale Cline connections more reliably.
- Unified Chat renders Cline CLI replies as readable messages instead of raw event output.
- Forgotten tab search can find hidden entries without restoring them to normal vault results.
SHIPPED · v1.5.0
v1.5.0 - Faster recall, forget controls, and Antigravity support
New
- Forget and Remember
- Antigravity CLI Support
- Faster Live Capture
Improved
- Much faster recall on large vaults — Search results now stay compact, show the relevant part of each memory, and avoid sending full saved conversations unless an AI tool asks for them
- Better memory ranking — Recent and frequently used memories get stronger placement, pinned memories remain first, and tag matches are now searchable
- Smarter related memory — Related-memory lookups and project-scoped injected context can include nearby memories that use different wording when semantic recall is enabled
Fixed
- Fixed **tag searches missing tagged-only memories** — entries now appear when the matching word is in a tag, even if it is not in the memory body
- Fixed **search previews starting at the beginning of long conversations** — previews now jump to the matching part of the memory so results are easier to choose
- Fixed **AI searches causing needless instruction file rewrites** — searching from connected clients no longer refreshes auto-injected files when no memory content changed
SHIPPED · v1.4.0
Auto-Inject Memory
New
- Auto-Inject Memory
- Claude Code Fresh Starts
Improved
- Safer instruction file updates — 1AIVault manages only its own memory block and preserves the rest of each AI tool's instruction file
- Hand-edited memory blocks are protected — If you manually change an injected block, 1AIVault skips the next rewrite instead of overwriting your edits
- Clearer injection status — Each tool row now shows whether memory is off, injected, handled by Claude Code's fresh-start option, or skipped because of a manual edit
Fixed
- Fixed **new memories saved from AI clients not reaching auto-injected context until restart** — memories saved through connected AI tools now refresh into enabled memory blocks while 1AIVault is running
SHIPPED · v1.3.2
Sharper, Faster Recall
Improved
- Much faster memory recall — Searching, browsing recent memories, opening a tag, and exploring related memories now complete in a fraction of the steps, so results come back noticeably quicker on larger vaults
- Smarter search ranking — A match in a memory's title now outranks the same word buried deep in its body, so the most relevant memory surfaces first
- Leaner search results — AI clients now receive a short preview of each matching memory and pull the full text only when they need it, keeping recall fast and responsive
Fixed
- Fixed **AI clients coming up empty when searching vaults with large saved conversations** — when a memory held a full saved conversation, search results grew too large and got cut off before the useful memories came through. Searches now return compact previews so the right memories always reach Claude Desktop, Claude Code, Cursor, and Cline — and the full text of any memory can still be pulled on demand
SHIPPED · v1.3.1
v1.3.1
Improved
- Far stronger OpenCode detection on Windows — Chat now checks Chocolatey, Scoop (system-wide), pnpm, nvm, mise, Yarn, Volta, and more install locations instead of relying on a short list of common paths
- Deep scan for hard-to-find installs — When OpenCode's binary lives inside a pnpm store or an unusual package layout, 1AIVault performs a bounded folder scan to find it automatically
- Smarter Windows command shims — pnpm-style and OpenCode-style shims that point to extension-less entry scripts are now resolved correctly, and a sibling `.exe` is used as a fallback for Scoop and WinGet layouts
Fixed
- Fixed **OpenCode not detected on Windows when installed via pnpm, Scoop, or Chocolatey** — every common Windows package manager layout is now checked
- Fixed **model picker silently hiding agents that failed to launch** — detected-but-broken agents now appear in a "Detected, but unavailable" section with the reason shown
- Fixed **model rescan not forcing a fresh agent detection** — refreshing models from the picker now triggers a full rescan of installed agents
SHIPPED · v1.3.0
Skills Hub & Unified Chat Agent Upgrades
New
- Skills Hub
- Unified Chat Agent Upgrades
Improved
- Stronger local agent detection — Chat now checks common install locations, remembers verified local AI tool paths, and rescans on demand across macOS, Windows, and Linux
- More useful Chat Context panel — Agent rows now show detected versions, locations, streaming support, errors, and managed OpenCode status
- Safer file and folder context — Drop, paste, type, or pick local paths in Chat and 1AIVault resolves them into reusable context chips
Fixed
- Fixed **Windows chat prompts failing with special characters** — Prompts containing symbols like `%`, `&`, or `|` are passed safely to local AI agents
- Fixed **Codex chat turns breaking on complex prompts** — Codex receives the prompt in a safer way, so longer instructions and pasted command text run more reliably
- Fixed **agent detection showing broken tools as ready** — Chat now reports setup errors earlier instead of waiting until the first message fails
SHIPPED · v1.2.5
Unified Chat crashing on Windows
Improved
- Double-click to classify — In the Classify Now wizard, double-click an AI tool to select it and jump straight to the next step
- Better wizard layout on small screens — Classify Now dialogs now scroll properly and stay centered, even on short windows or when the pro gate appears
Fixed
- Fixed **Unified Chat crashing on Windows when spawning a local AI agent** — Routing chat turns to Claude Code, Codex, Gemini CLI, OpenCode, Amp, Qwen, or Aider no longer fails with a cryptic error on Windows; the exit code is handled correctly so the conversation continues
7 older 1AIVault releases are listed in the release history.