Changelog

The meaningful changes in Nowledge Mem, written for people using it every day.

Latest release

0.10.91

2026년 9월 30일

0.10 is a ground-up engine rewrite

The 0.10.x line runs on Mem's rebuilt native engine, so most things are faster and lighter. It is a new engine, though, so you may still hit something unexpected. If you do, open Settings → Info, export your app.log, and send it to hello@nowledge-labs.ai and we'll get on it quickly. And if anything is blocking your work, you can download 0.9.29 from the Downloads page and roll back anytime.

0.10.95

Coming next

0.10.95NextComing next

Desktop Startup Recovery and More Reliable Search

Open Mem again after an update, find documents by their content, and recover unfinished search and conversation work.

Added

Ask AI Now for Mem help, change supported settings, and report a problem with available diagnostics and a screenshot in the same conversation.
Run an optional local HawDB self-test from Desktop Labs, with progress and useful failure details, without changing your existing data.
Ask AI Now to replace exact text in an existing task file, with a clear error when the requested match cannot be found.

Improved

Find documents by their indexed body and OCR text, including files outside the currently loaded Library list.
Make Fast decisions easier to find and configure, with clearer Local-mode guidance and readable model settings on narrow windows.
Use supported Jev models for Fast decisions through OpenRouter, Vercel AI Gateway, or Cloudflare AI Gateway in Vault App.
Choose an AI Identity before creating a new AI Now task on mobile.
Keep label selection usable with large label collections by searching and loading a bounded set of choices.
Use the native macOS menu bar icon treatment for consistent light and dark appearance.

Fixed

Open the desktop app after an update when its bundled backend version was incorrectly reported as 0.0.0.
Open the local workspace even when an internet request stalls during startup, instead of leaving an empty window.
Return to the conversation list when selecting Threads after opening sync settings, agent sessions, or a conversation.
Retry a failed search rebuild from its recorded stage and reuse verified remote embeddings, while text search remains available.
Resume safe unfinished chunks when retrying a long-conversation distillation, while retaining work already completed.
Keep time-aware search judgments that finish on time even when retrieval takes longer, with clearer notices when time interpretation is unavailable.
Keep every answer visible when reopening a newly created AI Now task after switching away from it.
Keep unsaved AI identity edits visible after a save failure and offer Retry, while preserving concurrent settings changes.
Apply the search-index disk-space reserve on Windows before rebuilding, helping protect nearly full drives.
Open the web app at /app with or without a trailing slash, including links with query parameters.
Correct entity aliases without deleting the entity or its links, and keep removed aliases excluded until you restore them.
Keep the selected Space when exporting a Thread, and preserve the original JSON, Markdown, or HTML format for legacy remote exports.
Give Devin lifecycle hooks a valid response even when capture fails, and report the reason for each failed session.

0.10.91

2026년 9월 30일

0.10.91Latest release2026년 9월 30일

More Reliable Spaces, Recovery, Search, and Connected Workflows

Make Space setup and connected workflows easier to follow, recover more safely from updates and local interruptions, and keep search results and useful AI work grounded in the right place.

Added

Discover supported schedule actions from connected AI Now connectors, so recurring work is easier to set up from the tools you already use.
Report a web page failure through Feedback with the available diagnostic context, so a broken view is easier to submit for review.

Improved

Use one clearer Space creation flow and carry the selected Space through hosted MCP setup, so a connection is less likely to land in the wrong place.
Keep ChatGPT model choices consistent between onboarding and Settings, and show Library AI replies as they stream instead of waiting for the full response.
Make failed macOS in-app updates easier to recover by keeping a usable app path and offering the latest matching installer when needed.
Honor an explicitly selected Space in local reads and preserve canonical Memory IDs in listings, so filtered views stay scoped to the place and records you chose.
Keep Tree activity details inside the preview pane, so reviewing an item does not unexpectedly move you away from the context you opened.

Fixed

Preserve canonical Skill bundle identity and recorded creator details, so saved and reviewed versions remain distinguishable.
Detect stale Cloud MCP registrations and make sync watcher failures visible and clearable instead of leaving a misleading healthy state.
Keep graph recovery evidence after an unclean shutdown and validate hosted conversation file imports before loading them.
Send browser users through the sign-in flow when authentication is required, instead of leaving them at a dead-end page.
Explain Feed errors with the diagnostic context needed for a useful report, so a broken feed is easier to investigate.
Retry transient source-repair timeouts within a bounded window, so a temporary search failure is less likely to interrupt a read.
Keep Cloud memory records bound to their canonical identity when aliases are involved, so reads and updates stay attached to the record you chose.
Keep offline vector generation non-empty for non-blank text, so local semantic search does not silently lose useful content.
Make ambiguous MCP feedback targets fail clearly, so feedback is attached to the item you intended.
Give Threads a clear Share menu and handoff links, so sharing a conversation takes fewer steps.
Treat an unsupported Library write destination as a hard error, so content cannot be stored in an unintended Space.
Show the scheduled execution lane for recurring work, so it is clear where an automated task will run.

0.10.89

2026년 9월 26일

0.10.892026년 9월 26일

Safer Updates, Clearer Decisions

Keep macOS updates recoverable, make Team Space migration state clear, and give you more confidence when reviewing Skills or choosing a Memory's Space.

Improved

Make macOS in-app updates more resilient when macOS denies the normal app-bundle swap, preserving a recoverable installed or staged app instead of leaving the app path missing.
Make a migrating Team Space's progress and completion readiness explicit, and include the required Space protocol acknowledgement in generated MCP configuration.
Show the recorded creator on same-name Skill review cards and in the Keep confirmation, so you can see which version you are keeping.

Fixed

Reject a Memory write to an unknown explicitly selected Space instead of silently falling back to the default Space.

0.10.88

2026년 9월 25일

0.10.882026년 9월 25일

Safer Recovery, Clearer Updates

Recover more local work with confidence, find the right installer when an update needs help, and keep everyday setup and account recovery easier to follow.

Added

Strengthen verified backup recovery for Graph, Content Store, feed, settings, and AI Now data with exact replay receipts, so Mem can account for every accepted source unit.

Improved

Keep Working Memory publication locks from being left behind after an interrupted publish, so a later publish can continue normally.
Show the authoritative monthly credit grants for Team workspaces, so account usage reflects the credits actually available to the team.
Make quota recovery easier to follow with clearer contrast and links that lead to the right next step.

Fixed

Make the official download the primary action after an update fails, so you can quickly recover with the matching installer.
Guide versioned remote OpenCode installs to the right setup instructions for the host you are using.
Keep the macOS DMG installation layout clear in Finder, with the app and Applications folder arranged for the expected drag-and-drop install.

0.10.87

2026년 9월 25일

0.10.872026년 9월 25일

More Reliable Workflows, Better Recovery

Bring documents and Working Memory into Mem more safely, keep Cloud background work and timelines dependable, and make recovery clearer across connected tools.

Added

Let connected MCP agents add text, PDF, and Office documents directly to your Library, with the same parsing, deduplication, labels, and access controls as a manual upload.
Try HawDB Labs with a clearer first-enable flow that prepares the experimental local workspace without replacing your existing graph.
Import a verified Working Memory ZIP into Cloud when its manifest, files, hashes, and counts match exactly; unsupported archive families are rejected instead of silently dropping data.
Manage feature flags from the operator dashboard with server-backed creation, rollouts, and overrides.

Improved

Keep a Message search rebuild safely owned when its native index work takes longer than the caller wait, so Mem avoids overlapping rebuilds and can settle a late result correctly.
Use a fresh, account-specific ChatGPT Subscription model catalog before saving a model choice, so unavailable or stale models are not selected by mistake.
Keep Cloud background review batches on one durable run, so a retry resumes only the remaining items without duplicating completed work or inflating usage.
Keep the activity timeline available while a feed read is still pending, instead of making the whole view wait.
Distinguish same-name Skills in "Worth a look" by showing their creators when that identity is available.

Fixed

Keep streamed AI Now text visible when a connection ends before the final response event, so useful partial replies are not lost.
Keep browser-extension authentication recovery visible before a Thread sync and even when no conversation is open, with a direct reconnect path instead of a misleading healthy state.
Install valid signed macOS updates more reliably, preserve detailed recovery errors, and keep the DMG installation guide visible in Finder.
Preserve local graph recovery files after an unclean exit so Mem can safely replay or diagnose the database instead of discarding evidence.

0.10.86

2026년 9월 24일

0.10.862026년 9월 24일

A Clearer, Safer Workspace

Choose a ChatGPT model with confidence, keep imports truthful, retain PDF text, and make local storage and setup easier to understand.

Added

Choose GPT-6 Astra, Sol, or Luna for a connected ChatGPT Subscription when it is available to your account, and keep that choice after you save and reopen Settings.
Set up the browser extension in its own onboarding step, so connecting it no longer competes with your other first-run choices.

Improved

Navigate shared controls and the welcome guide more comfortably with visible keyboard focus, arrow-key choices, named icon actions, and predictable focus when a guide closes.
Understand HawDB Labs before you opt in: Mem explains that it is experimental and local, shows verified restart results, and lets you recheck readiness without restarting again.

Fixed

Onboarding imports keep finished sources and label any remaining sources as incomplete, rather than declaring a partial run successful.
Keep native text available while processing PDFs, so readable pages do not lose their text when visual extraction is also needed.
Protect your local graph when its data root cannot be trusted: Mem fails safely instead of reusing an uncertain location and reports when graph memory is running in a degraded state.
Complete ChatGPT Subscription browser sign-in more reliably and keep saved credentials through a temporary request failure, so a recoverable interruption does not force a reconnect.
Keep your first onboarding question draftable while imports or provider setup finish, while still waiting to send it until setup is ready.
Turn failed views into clear recovery actions instead of a dead end.
Let a configured Mem Vault account continue through onboarding without being blocked by managed-AI setup intended for a different mode.

0.10.85

2026년 9월 22일

0.10.852026년 9월 22일

Keep Work Recoverable

Back up Working Memory exactly, finish folder imports, spend background budget in money, and recover drafts after resume.

Highlights

Back up Working Memory as exact facts, not a best-effort export.
See background work in money, and know when a safety rail paused it.
Finish folder imports and keep Memory drafts after you step away.
Keep queued work and transcripts on the account you signed in with.

Added

Connect Step Code natively: start with your current Mem context, keep completed coding conversations in Threads, and import older Step Code sessions when you choose.
Keep exact Working Memory facts in a backup and reject a malformed bundle before anything is overwritten.
See App background work as money, including the rolling daily window and a clear pause when a safety rail stops paid work.
Copy an agent continuation so you can pick up the same thread later.
Choose Fast decisions as a model purpose, including the TypeSafe provider, when you want a bounded judgment instead of a long chat.

Improved

Browse local coding sessions in pages instead of loading every session at once.
Keep a Memory draft when you close and reopen the prompt.
Make generated output follow the app language.
Keep AI Now task controls usable on a phone-sized layout.
Restart local search after you change the embedding model, instead of keeping the previous runtime.

Fixed

Finish a folder import without duplicate retries, and account for every file in that folder.
Ask which AI provider you want before onboarding asks for an API key, and open the matching browser extension store.
Restore the search rebuild action after a rebuild fails, and resume oversized PDF pages in smaller regions instead of failing the whole page.
Keep queued work and AI Now transcripts on the signed-in account, and restore connector state after the app resumes.
Hide health alerts for conversation sources you have turned off, and revoke client credentials when you sign out.
Give long thread reviews more budget, and recover when structured background output is cut short.

0.10.84

2026년 9월 21일

0.10.842026년 9월 21일

Stay in Context

Keep the current Space stable, search while background work runs, and bring more coding sessions into Mem

Highlights

Search without waiting for document ingestion to finish.
Bring Cherry Studio and Cursor coding history into Mem.
Keep the current Space stable while you explore other work.

Added

Sync conversations from Cherry Studio as a local session source.
Import Cursor composer history from its local VS Code database.
Install Mem on macOS with a clear drag-to-Applications guide in the DMG.

Improved

Preview or manage another Space without redirecting where later work is saved.
Keep long fenced code inside AI Now messages, with readable scrolling and unchanged copied text.
Return local search results promptly while documents are being embedded in the background.
Keep the selected project visible when switching between supported local coding clients.
Set up OpenCode against its v2 connection contract with clearer recovery when setup needs attention.
Make Stats follow the Space you are currently viewing.
Accept common Google AI Studio API base URLs when configuring Gemini.

Fixed

Open the exact Memory a search result refers to, even when similarly named records exist.
Keep search rebuilds from accumulating an unbounded number of old message versions.
Settle Memory status flags correctly after a search rebuild completes.
Start a loopback-authenticated local backend without treating its protected health check as a failure.
Restore connected sessions after the local backend restarts.
Protect Codex session imports when concurrent configuration updates would otherwise conflict.
Repair split OpenCode histories without duplicating conversation content.
Keep the Timeline at a usable width on tablet-sized layouts.
Use consistent, centered loading indicators across the app.
Complete Grok Bot sign-in through its native authorization callback.
Keep desktop sidebar labels readable and consistently aligned.
Recover a timed-out Cloud sign-in without losing the secure handoff.

0.10.83

2026년 9월 19일

0.10.832026년 9월 19일

Keep Work Moving

Recover conversation capture safely, follow background work clearly, and keep search and connections moving.

Highlights

Recover Codex conversation capture without losing new turns.
Browse local coding-agent sessions from Connectors.
See exactly what needs attention when search repair runs.

Added

Browse supported local coding-agent sessions directly from Connectors, then choose exactly what to import.
See the provider and model that handled each Processing task, including useful fallback and retry context.
Follow background-task progress and execution health instead of being left with an unexplained working state.

Improved

Import supported compressed Codex session snapshots.
Search repair names the affected conversations, Sources, and Memories and offers the right recovery action for each.
Watch durable reindex write progress, so a long rebuild visibly keeps moving.
Keep large Source repair results bounded, so recovery views stay responsive.
Return to the same workspace when you reopen or retry supported agent-session discovery.
See why one overlapping rule takes priority over another.
Keep preferred vocabulary during distillation review and leave syntax noise out of labels.
Search canonical Memory and Wiki documents through KFS.
Separate a skill's explicit unavailable result from relevance, so recommendations are easier to interpret.
Use the current GPT-6.5 catalog when selecting a ChatGPT Subscription model.
Restore a pending device login when you reopen the app.
Keep local projection settings available when you're offline.

Fixed

Codex capture recovers when a saved conversation already exists, without dropping a genuinely new repeated message.
ChatGPT Subscription work recovers from ordinary sign-in expiry and tells you when reconnecting is needed.
Memory extraction waits for PDF processing to finish, so a document is not distilled before its content is ready.
Same-Wi-Fi access now keeps its promised API-key protection, while ordinary localhost use remains unchanged.
Desktop startup checks only the configured backend address, so an unrelated listener no longer blocks a local Mem backend.
Working Memory rechecks your current permission and preserves data through interrupted recovery.
Manual Knowledge Graph repairs keep their admission state across interruption and cancellation.
Background task failures offer recovery instead of leaving work stuck.
Keep partial Source health visible after a retry.
Reject unsupported agent transcript formats instead of importing unreliable data.
Local imports retain parallel relationships and Space IDs.
Recover the AI Now composer after an interrupted turn.
Keep an uploaded Source visible after Library refresh.
Show OCR progress for PDF Sources.
Keep message capture independent from AI turns.

0.10.82

2026년 9월 15일

0.10.822026년 9월 15일

Keep Every Detail

Read complete knowledge, revisit document history, and connect your AI tools to the right Mem.

Added

Adjust search wait times and retries in Settings when your embedding provider responds slowly.
Import conversations exported from Kagi Assistant as Markdown.

Improved

Browse a document's complete version history without older versions crowding the Library list.
Connector setup now uses the local or remote Mem selected in your App and guides your AI tool through connection checks.
Desktop keeps checking for updates while its window is hidden, and small graph previews pause when out of view.
Space summaries distinguish active Memories from archived items and Crystals.

Fixed

Review proposed Memory changes and their supporting evidence before confirming a Timeline decision.
Memory history corrections show the full contents being reviewed and ask you to refresh if the evidence changes.
Conversation deletion reports completed work and unfinished cleanup, keeping failed items available for follow-up.
AI Now preserves nested lists, indented code, and Markdown line breaks in model replies.
OAuth connector setup completes correctly when the provider redirects its callback address.
Background AI work respects local-only mode instead of unexpectedly contacting a remote model.
Saving an older Memory policy draft now reports a conflict instead of overwriting a newer change.
Label edits stay consistent across open lists, including recovery after a failed save.
Codex session capture preserves image attachments from current conversation records.
Background jobs stop repeating an output-budget retry when it cannot change the outcome.
Background work exposes actionable failures and bounded progress instead of remaining stuck on an unexplained working state.
Memory reviews whose pending state was left behind are reconciled so they do not remain stuck indefinitely.
AI tools receive the full text when opening a Memory, and complete conversation exports retain messages beyond the first page.
Resuming PDF OCR preserves the document's original text, continues through all remaining pages, and reuses work already completed.
Deleting a Source also removes its stored content and prevents ongoing OCR from bringing it back.
Re-uploading a coding session to an explicitly chosen Space moves the existing conversation there without creating a duplicate.
Session lifecycle capture respects your auto-sync preference, and local MCP attachment saving no longer reports a remote-only restriction.
Space selection remains consistent across the App, CLI, and AI tools, including Rules and Ontologies.
Crystal evidence loading failures show a retry action, and interrupted remote live updates are clearly identified.
Open a Memory from a standalone graph in the desktop app or its web fallback.
Re-importing an unchanged Skill bundle no longer creates a redundant update.

Ready to try Mem?

Download the app, connect your AI tools, and keep useful context where you control it.

Download Read Docs

Changelog

The meaningful changes in Nowledge Mem, written for people using it every day.

Latest release

0.10.91

2026년 9월 30일

0.10 is a ground-up engine rewrite

0.10.95

Coming next

0.10.95NextComing next

Desktop Startup Recovery and More Reliable Search

Open Mem again after an update, find documents by their content, and recover unfinished search and conversation work.

Added

Ask AI Now for Mem help, change supported settings, and report a problem with available diagnostics and a screenshot in the same conversation.
Run an optional local HawDB self-test from Desktop Labs, with progress and useful failure details, without changing your existing data.
Ask AI Now to replace exact text in an existing task file, with a clear error when the requested match cannot be found.

Improved

Find documents by their indexed body and OCR text, including files outside the currently loaded Library list.
Make Fast decisions easier to find and configure, with clearer Local-mode guidance and readable model settings on narrow windows.
Use supported Jev models for Fast decisions through OpenRouter, Vercel AI Gateway, or Cloudflare AI Gateway in Vault App.
Choose an AI Identity before creating a new AI Now task on mobile.
Keep label selection usable with large label collections by searching and loading a bounded set of choices.
Use the native macOS menu bar icon treatment for consistent light and dark appearance.

Fixed

Open the desktop app after an update when its bundled backend version was incorrectly reported as 0.0.0.
Open the local workspace even when an internet request stalls during startup, instead of leaving an empty window.
Return to the conversation list when selecting Threads after opening sync settings, agent sessions, or a conversation.
Retry a failed search rebuild from its recorded stage and reuse verified remote embeddings, while text search remains available.
Resume safe unfinished chunks when retrying a long-conversation distillation, while retaining work already completed.
Keep time-aware search judgments that finish on time even when retrieval takes longer, with clearer notices when time interpretation is unavailable.
Keep every answer visible when reopening a newly created AI Now task after switching away from it.
Keep unsaved AI identity edits visible after a save failure and offer Retry, while preserving concurrent settings changes.
Apply the search-index disk-space reserve on Windows before rebuilding, helping protect nearly full drives.
Open the web app at /app with or without a trailing slash, including links with query parameters.
Correct entity aliases without deleting the entity or its links, and keep removed aliases excluded until you restore them.
Keep the selected Space when exporting a Thread, and preserve the original JSON, Markdown, or HTML format for legacy remote exports.
Give Devin lifecycle hooks a valid response even when capture fails, and report the reason for each failed session.

0.10.91

2026년 9월 30일

0.10.91Latest release2026년 9월 30일

More Reliable Spaces, Recovery, Search, and Connected Workflows

Make Space setup and connected workflows easier to follow, recover more safely from updates and local interruptions, and keep search results and useful AI work grounded in the right place.

Added

Discover supported schedule actions from connected AI Now connectors, so recurring work is easier to set up from the tools you already use.
Report a web page failure through Feedback with the available diagnostic context, so a broken view is easier to submit for review.

Improved

Use one clearer Space creation flow and carry the selected Space through hosted MCP setup, so a connection is less likely to land in the wrong place.
Keep ChatGPT model choices consistent between onboarding and Settings, and show Library AI replies as they stream instead of waiting for the full response.
Make failed macOS in-app updates easier to recover by keeping a usable app path and offering the latest matching installer when needed.
Honor an explicitly selected Space in local reads and preserve canonical Memory IDs in listings, so filtered views stay scoped to the place and records you chose.
Keep Tree activity details inside the preview pane, so reviewing an item does not unexpectedly move you away from the context you opened.

Fixed

Preserve canonical Skill bundle identity and recorded creator details, so saved and reviewed versions remain distinguishable.
Detect stale Cloud MCP registrations and make sync watcher failures visible and clearable instead of leaving a misleading healthy state.
Keep graph recovery evidence after an unclean shutdown and validate hosted conversation file imports before loading them.
Send browser users through the sign-in flow when authentication is required, instead of leaving them at a dead-end page.
Explain Feed errors with the diagnostic context needed for a useful report, so a broken feed is easier to investigate.
Retry transient source-repair timeouts within a bounded window, so a temporary search failure is less likely to interrupt a read.
Keep Cloud memory records bound to their canonical identity when aliases are involved, so reads and updates stay attached to the record you chose.
Keep offline vector generation non-empty for non-blank text, so local semantic search does not silently lose useful content.
Make ambiguous MCP feedback targets fail clearly, so feedback is attached to the item you intended.
Give Threads a clear Share menu and handoff links, so sharing a conversation takes fewer steps.
Treat an unsupported Library write destination as a hard error, so content cannot be stored in an unintended Space.
Show the scheduled execution lane for recurring work, so it is clear where an automated task will run.

0.10.89

2026년 9월 26일

0.10.892026년 9월 26일

Safer Updates, Clearer Decisions

Keep macOS updates recoverable, make Team Space migration state clear, and give you more confidence when reviewing Skills or choosing a Memory's Space.

Improved

Make macOS in-app updates more resilient when macOS denies the normal app-bundle swap, preserving a recoverable installed or staged app instead of leaving the app path missing.
Make a migrating Team Space's progress and completion readiness explicit, and include the required Space protocol acknowledgement in generated MCP configuration.
Show the recorded creator on same-name Skill review cards and in the Keep confirmation, so you can see which version you are keeping.

Fixed

Reject a Memory write to an unknown explicitly selected Space instead of silently falling back to the default Space.

0.10.88

2026년 9월 25일

0.10.882026년 9월 25일

Safer Recovery, Clearer Updates

Recover more local work with confidence, find the right installer when an update needs help, and keep everyday setup and account recovery easier to follow.

Added

Strengthen verified backup recovery for Graph, Content Store, feed, settings, and AI Now data with exact replay receipts, so Mem can account for every accepted source unit.

Improved

Keep Working Memory publication locks from being left behind after an interrupted publish, so a later publish can continue normally.
Show the authoritative monthly credit grants for Team workspaces, so account usage reflects the credits actually available to the team.
Make quota recovery easier to follow with clearer contrast and links that lead to the right next step.

Fixed

Make the official download the primary action after an update fails, so you can quickly recover with the matching installer.
Guide versioned remote OpenCode installs to the right setup instructions for the host you are using.
Keep the macOS DMG installation layout clear in Finder, with the app and Applications folder arranged for the expected drag-and-drop install.

0.10.87

2026년 9월 25일

0.10.872026년 9월 25일

More Reliable Workflows, Better Recovery

Bring documents and Working Memory into Mem more safely, keep Cloud background work and timelines dependable, and make recovery clearer across connected tools.

Added

Let connected MCP agents add text, PDF, and Office documents directly to your Library, with the same parsing, deduplication, labels, and access controls as a manual upload.
Try HawDB Labs with a clearer first-enable flow that prepares the experimental local workspace without replacing your existing graph.
Import a verified Working Memory ZIP into Cloud when its manifest, files, hashes, and counts match exactly; unsupported archive families are rejected instead of silently dropping data.
Manage feature flags from the operator dashboard with server-backed creation, rollouts, and overrides.

Improved

Keep a Message search rebuild safely owned when its native index work takes longer than the caller wait, so Mem avoids overlapping rebuilds and can settle a late result correctly.
Use a fresh, account-specific ChatGPT Subscription model catalog before saving a model choice, so unavailable or stale models are not selected by mistake.
Keep Cloud background review batches on one durable run, so a retry resumes only the remaining items without duplicating completed work or inflating usage.
Keep the activity timeline available while a feed read is still pending, instead of making the whole view wait.
Distinguish same-name Skills in "Worth a look" by showing their creators when that identity is available.

Fixed

Keep streamed AI Now text visible when a connection ends before the final response event, so useful partial replies are not lost.
Keep browser-extension authentication recovery visible before a Thread sync and even when no conversation is open, with a direct reconnect path instead of a misleading healthy state.
Install valid signed macOS updates more reliably, preserve detailed recovery errors, and keep the DMG installation guide visible in Finder.
Preserve local graph recovery files after an unclean exit so Mem can safely replay or diagnose the database instead of discarding evidence.

0.10.86

2026년 9월 24일

0.10.862026년 9월 24일

A Clearer, Safer Workspace

Choose a ChatGPT model with confidence, keep imports truthful, retain PDF text, and make local storage and setup easier to understand.

Added

Choose GPT-6 Astra, Sol, or Luna for a connected ChatGPT Subscription when it is available to your account, and keep that choice after you save and reopen Settings.
Set up the browser extension in its own onboarding step, so connecting it no longer competes with your other first-run choices.

Improved

Navigate shared controls and the welcome guide more comfortably with visible keyboard focus, arrow-key choices, named icon actions, and predictable focus when a guide closes.
Understand HawDB Labs before you opt in: Mem explains that it is experimental and local, shows verified restart results, and lets you recheck readiness without restarting again.

Fixed

Onboarding imports keep finished sources and label any remaining sources as incomplete, rather than declaring a partial run successful.
Keep native text available while processing PDFs, so readable pages do not lose their text when visual extraction is also needed.
Protect your local graph when its data root cannot be trusted: Mem fails safely instead of reusing an uncertain location and reports when graph memory is running in a degraded state.
Complete ChatGPT Subscription browser sign-in more reliably and keep saved credentials through a temporary request failure, so a recoverable interruption does not force a reconnect.
Keep your first onboarding question draftable while imports or provider setup finish, while still waiting to send it until setup is ready.
Turn failed views into clear recovery actions instead of a dead end.
Let a configured Mem Vault account continue through onboarding without being blocked by managed-AI setup intended for a different mode.

0.10.85

2026년 9월 22일

0.10.852026년 9월 22일

Keep Work Recoverable

Back up Working Memory exactly, finish folder imports, spend background budget in money, and recover drafts after resume.

Highlights

Back up Working Memory as exact facts, not a best-effort export.
See background work in money, and know when a safety rail paused it.
Finish folder imports and keep Memory drafts after you step away.
Keep queued work and transcripts on the account you signed in with.

Added

Connect Step Code natively: start with your current Mem context, keep completed coding conversations in Threads, and import older Step Code sessions when you choose.
Keep exact Working Memory facts in a backup and reject a malformed bundle before anything is overwritten.
See App background work as money, including the rolling daily window and a clear pause when a safety rail stops paid work.
Copy an agent continuation so you can pick up the same thread later.
Choose Fast decisions as a model purpose, including the TypeSafe provider, when you want a bounded judgment instead of a long chat.

Improved

Browse local coding sessions in pages instead of loading every session at once.
Keep a Memory draft when you close and reopen the prompt.
Make generated output follow the app language.
Keep AI Now task controls usable on a phone-sized layout.
Restart local search after you change the embedding model, instead of keeping the previous runtime.

Fixed

Finish a folder import without duplicate retries, and account for every file in that folder.
Ask which AI provider you want before onboarding asks for an API key, and open the matching browser extension store.
Restore the search rebuild action after a rebuild fails, and resume oversized PDF pages in smaller regions instead of failing the whole page.
Keep queued work and AI Now transcripts on the signed-in account, and restore connector state after the app resumes.
Hide health alerts for conversation sources you have turned off, and revoke client credentials when you sign out.
Give long thread reviews more budget, and recover when structured background output is cut short.

0.10.84

2026년 9월 21일

0.10.842026년 9월 21일

Stay in Context

Keep the current Space stable, search while background work runs, and bring more coding sessions into Mem

Highlights

Search without waiting for document ingestion to finish.
Bring Cherry Studio and Cursor coding history into Mem.
Keep the current Space stable while you explore other work.

Added

Sync conversations from Cherry Studio as a local session source.
Import Cursor composer history from its local VS Code database.
Install Mem on macOS with a clear drag-to-Applications guide in the DMG.

Improved

Preview or manage another Space without redirecting where later work is saved.
Keep long fenced code inside AI Now messages, with readable scrolling and unchanged copied text.
Return local search results promptly while documents are being embedded in the background.
Keep the selected project visible when switching between supported local coding clients.
Set up OpenCode against its v2 connection contract with clearer recovery when setup needs attention.
Make Stats follow the Space you are currently viewing.
Accept common Google AI Studio API base URLs when configuring Gemini.

Fixed

Open the exact Memory a search result refers to, even when similarly named records exist.
Keep search rebuilds from accumulating an unbounded number of old message versions.
Settle Memory status flags correctly after a search rebuild completes.
Start a loopback-authenticated local backend without treating its protected health check as a failure.
Restore connected sessions after the local backend restarts.
Protect Codex session imports when concurrent configuration updates would otherwise conflict.
Repair split OpenCode histories without duplicating conversation content.
Keep the Timeline at a usable width on tablet-sized layouts.
Use consistent, centered loading indicators across the app.
Complete Grok Bot sign-in through its native authorization callback.
Keep desktop sidebar labels readable and consistently aligned.
Recover a timed-out Cloud sign-in without losing the secure handoff.

0.10.83

2026년 9월 19일

0.10.832026년 9월 19일

Keep Work Moving

Recover conversation capture safely, follow background work clearly, and keep search and connections moving.

Highlights

Recover Codex conversation capture without losing new turns.
Browse local coding-agent sessions from Connectors.
See exactly what needs attention when search repair runs.

Added

Browse supported local coding-agent sessions directly from Connectors, then choose exactly what to import.
See the provider and model that handled each Processing task, including useful fallback and retry context.
Follow background-task progress and execution health instead of being left with an unexplained working state.

Improved

Import supported compressed Codex session snapshots.
Search repair names the affected conversations, Sources, and Memories and offers the right recovery action for each.
Watch durable reindex write progress, so a long rebuild visibly keeps moving.
Keep large Source repair results bounded, so recovery views stay responsive.
Return to the same workspace when you reopen or retry supported agent-session discovery.
See why one overlapping rule takes priority over another.
Keep preferred vocabulary during distillation review and leave syntax noise out of labels.
Search canonical Memory and Wiki documents through KFS.
Separate a skill's explicit unavailable result from relevance, so recommendations are easier to interpret.
Use the current GPT-6.5 catalog when selecting a ChatGPT Subscription model.
Restore a pending device login when you reopen the app.
Keep local projection settings available when you're offline.

Fixed

Codex capture recovers when a saved conversation already exists, without dropping a genuinely new repeated message.
ChatGPT Subscription work recovers from ordinary sign-in expiry and tells you when reconnecting is needed.
Memory extraction waits for PDF processing to finish, so a document is not distilled before its content is ready.
Same-Wi-Fi access now keeps its promised API-key protection, while ordinary localhost use remains unchanged.
Desktop startup checks only the configured backend address, so an unrelated listener no longer blocks a local Mem backend.
Working Memory rechecks your current permission and preserves data through interrupted recovery.
Manual Knowledge Graph repairs keep their admission state across interruption and cancellation.
Background task failures offer recovery instead of leaving work stuck.
Keep partial Source health visible after a retry.
Reject unsupported agent transcript formats instead of importing unreliable data.
Local imports retain parallel relationships and Space IDs.
Recover the AI Now composer after an interrupted turn.
Keep an uploaded Source visible after Library refresh.
Show OCR progress for PDF Sources.
Keep message capture independent from AI turns.

0.10.82

2026년 9월 15일

0.10.822026년 9월 15일

Keep Every Detail

Read complete knowledge, revisit document history, and connect your AI tools to the right Mem.

Added

Adjust search wait times and retries in Settings when your embedding provider responds slowly.
Import conversations exported from Kagi Assistant as Markdown.

Improved

Browse a document's complete version history without older versions crowding the Library list.
Connector setup now uses the local or remote Mem selected in your App and guides your AI tool through connection checks.
Desktop keeps checking for updates while its window is hidden, and small graph previews pause when out of view.
Space summaries distinguish active Memories from archived items and Crystals.

Fixed

Review proposed Memory changes and their supporting evidence before confirming a Timeline decision.
Memory history corrections show the full contents being reviewed and ask you to refresh if the evidence changes.
Conversation deletion reports completed work and unfinished cleanup, keeping failed items available for follow-up.
AI Now preserves nested lists, indented code, and Markdown line breaks in model replies.
OAuth connector setup completes correctly when the provider redirects its callback address.
Background AI work respects local-only mode instead of unexpectedly contacting a remote model.
Saving an older Memory policy draft now reports a conflict instead of overwriting a newer change.
Label edits stay consistent across open lists, including recovery after a failed save.
Codex session capture preserves image attachments from current conversation records.
Background jobs stop repeating an output-budget retry when it cannot change the outcome.
Background work exposes actionable failures and bounded progress instead of remaining stuck on an unexplained working state.
Memory reviews whose pending state was left behind are reconciled so they do not remain stuck indefinitely.
AI tools receive the full text when opening a Memory, and complete conversation exports retain messages beyond the first page.
Resuming PDF OCR preserves the document's original text, continues through all remaining pages, and reuses work already completed.
Deleting a Source also removes its stored content and prevents ongoing OCR from bringing it back.
Re-uploading a coding session to an explicitly chosen Space moves the existing conversation there without creating a duplicate.
Session lifecycle capture respects your auto-sync preference, and local MCP attachment saving no longer reports a remote-only restriction.
Space selection remains consistent across the App, CLI, and AI tools, including Rules and Ontologies.
Crystal evidence loading failures show a retry action, and interrupted remote live updates are clearly identified.
Open a Memory from a standalone graph in the desktop app or its web fallback.
Re-importing an unchanged Skill bundle no longer creates a redundant update.

Ready to try Mem?

Download the app, connect your AI tools, and keep useful context where you control it.

Download Read Docs