Changelog

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

Latest release

0.10.62

Aug 13, 2026

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.

0.10.62

Aug 13, 2026

0.10.62Latest releaseAug 13, 2026

DeepSeek Harness, ZCode, and Safer Backups

Highlights

DeepSeek Harness and ZCode are now first-class Mem integrations.
Backups restore Library original files, not just parsed text.
Long remote-embedding rebuilds are bounded before they reach your provider.

Added

Fixed

Cloud restore now brings Library originals back with the backup. PDFs, documents, images, and other uploaded files can be previewed or downloaded after restore instead of leaving only parsed text behind.
Search Index rebuilds no longer send oversized Memory or Library text straight to a remote embedding provider. Long items are bounded before the request, so one large note is much less likely to keep semantic search stuck in text-only mode.
ChatGPT custom MCP connector setup is more reliable when OAuth discovery is probed after Allow. The pairing page now answers the compatibility endpoint without widening the connector's default read-only scope.
Data Transfer and Library recovery now fail more explicitly: recoverable export and restore mismatches become warnings, while binary-looking Library uploads are quarantined instead of being indexed as broken placeholder text.
AI Now no longer automatically replays slow or interrupted model requests. If a local or self-hosted model is still working, Mem now stops and lets you continue deliberately instead of queuing the same large prompt again.
The Threads source filter now treats case variants as the same source. Sessions from Hermes, hermes, and related Hermes agent names appear under one clean source instead of duplicate filters.
Desktop installers and self-hosted Docker images are now built with a more resilient PDF parser dependency download, so brief network interruptions during release packaging are less likely to delay an update.

cloud-2026-08-13

Aug 13, 2026

cloud-2026-08-13Aug 13, 2026

A Better Home for Teams on Nowledge Cloud

A wave of improvements for team workspaces on Nowledge Cloud. They arrive with your workspace, so there is nothing to install or update.

Highlights

Every Mem you belong to is one switch away, with Personal and Team clearly told apart.
Team imports explain who can read what before you commit.

Added

See every Mem you belong to and switch between them. Workspace pickers now tell Personal from Team at a glance, so you always know whose memory you are looking at.
Name your workspace. Rename it right from the switcher, and a new team is invited to pick its name the first time it opens.
Plan and seats in one place. Your team's plan, seat usage, and a direct path to billing now live together on the Team page.

Improved

Signing in takes you to your team first, so shared work is one step closer.
Application and sign-in emails now arrive in your language and address your company by name.
Invites tell the truth. If an invite email could not be delivered, Mem says so clearly, instead of leaving a teammate waiting on a message that never comes.
Importing into a team workspace now explains up front who will be able to read what, before anything is brought in.

0.10.59

Aug 13, 2026

0.10.59Aug 13, 2026

Clearer Setup and Graph Feedback

Highlights

Graph Explorer no longer gets stuck on an empty load.
Distillation tells you when a language model is missing.
Update install stays explicitly user-confirmed.

Fixed

Graph Explorer inside MCP-capable clients no longer sits on an empty loading screen. If the host opens the graph before sending data, Mem now tells you the next action to run.
Graph Intelligence is more reliable when you ask about the selected item. Mem now keeps the selected graph object in the chat context even when the canvas sends it through a wrapped event shape.
Conversation distillation now explains the real fix when a language model is missing. Downloading the Search Index model is not enough for distillation; Mem points you to the Default or Agents language model setting instead of showing a raw 422 error.
The update action is clearer after a download finishes. The title-bar update button now asks for the same explicit install confirmation as Settings, and Mem still never installs or restarts from a background download without your action.

Improved

Vision settings now say what actually happens in every language: if you do not choose a remote vision provider, Library uses the on-device local vision model rather than inheriting your default text model.
Long-running Codex and AI Now search cards stay compact while they are still working, then open when results are ready, so a single tool call no longer takes over the whole conversation.
Busy graph canvases get a quieter ontology legend that stays compact until you expand it or interact with a type.

0.10.57

Aug 12, 2026

0.10.57Aug 12, 2026

Connector and Library Reliability

Highlights

ChatGPT connector setup works through Access Anywhere again.
Oversized Library documents are bounded before embedding.
Kimi Code gets a smaller connector schema.

Fixed

ChatGPT custom MCP connector setup can complete through Access Anywhere again. After you click Allow, Mem now lets the authenticated setup flow read the status and pairing endpoints it needs, while still blocking public management access.
Library indexing is stricter about bad or oversized documents. Very long markdown, docx-derived text, and unusual source content are split and bounded before embedding, so one broken item should not take the whole refresh down or overload your provider's context limit.
All Spaces in Library now opens a source with the source's real Space behind it, so parsed content and details do not disappear just because you were browsing the aggregate view.
Windows launch-at-login is cleaned up at startup when the setting is off, so an old OS auto-start registration no longer brings Mem back after you disabled it.
Backups can now continue when an old Thread is missing historical message payloads. Mem records the missing content as a warning in the export manifest instead of blocking the rest of your data from being saved.

Improved

Kimi Code now gets the compact read-only connector profile instead of the full app tool surface, keeping setup lighter and tool schemas smaller.
Update downloads avoid cellular connections even when background downloading is otherwise allowed. Mem still downloads only in the background and never installs or restarts without your action.

0.10.56

Aug 11, 2026

0.10.56Aug 11, 2026

Memory That Keeps Growing

Fixed

Mem keeps learning as you use it more. Some of you told us your memory count stopped moving for days, in one case about a week, while you were working hard every day and everything looked switched on. Background learning had drifted apart from the part that actually writes a memory down, so the app could look busy without learning anything. It is one path again, and it is checked.
Cleaning up is something you decide, not something that happens to you. Maintenance now shows you what it wants to review before anything moves, so a number on a card is no longer a thing you have to trust blindly. Nothing is thrown away quietly.
Wiki topics refresh again, and their summaries follow your newer reading instead of stopping at what you knew months ago. The graph is quieter too: fewer orphan and near-duplicate entities cluttering what you are trying to read.
Search puts trustworthy memories first and does it before deciding which page you are on, so a good result can no longer fall off the end. Date filters on Memories work as you would expect.
AI Now is steadier: tools with unusual inputs no longer get refused, moving between tasks keeps its place, an idle session says it is idle, and a subagent uses the model you picked for the session.
A backup that could not capture a conversation completely is refused instead of written, so a restore does not hand you a half-finished thread.

Added

Share to Mem from your phone. Anything you can share on iOS can go straight into your Library without opening the app first.
Cindy and Mirasim join the connectors, and OpenCode works on Windows.

0.10.55

Aug 9, 2026

0.10.55Aug 9, 2026

Every Agent You Already Use

Added

Session sync now finds conversations from every AI agent Mem knows how to read, not just the five it used to look for. If you work in Hermes, Gemini CLI, Kimi, Grok, Antigravity or Pi, those conversations are waiting when you open session sync, and you choose which are worth keeping.
Mem now ships a portable plugin for tools that follow the Agent Plugins standard but do not have a dedicated Nowledge connector yet. It gives that tool search, memory saving, and handoff summaries. Where a dedicated connector exists, keep using it: it does more.
Amp now has a native Nowledge Mem connector. New Amp sessions can start with your Context Bundle, use memory and graph tools directly, and save real Amp threads automatically after agent turns.
Google Antigravity now has a dedicated Mem plugin, thanks to abn's contribution. Antigravity sessions can start with your current context, use Mem MCP tools and skills directly, and save real local transcripts automatically when a session ends.
Ekko Agent in Hermes Studio can now bring its exact local conversations into Mem. Ekko still uses MCP for context and memory tools; the new `nmem t sync --from ekko` importer reads completed runs from Ekko's SQLite message store and saves them as searchable `source=ekko` Threads.
Mirasim joins the connector guide. Add Mem as a Mirasim MCP server for the built-in agent, bring in the shared skills when your build supports them, and keep Claude Code or Codex panes on their dedicated connectors for exact thread capture.
Cindy is now in the connector guide too. Cindy's built-in agent can use Mem through MCP and shared skills; sessions that Cindy runs through Codex, Claude Code, Pi, or OMP keep their dedicated connectors for exact transcript capture.

Fixed

Refreshing search finishes, or tells you it failed, instead of running all night. A rebuild leaning on a remote provider that stopped answering could keep working for hours with nothing to show; it now gives up on a clear deadline. A rebuild wrongly reported as stalled can recover and finish rather than staying stuck on screen.
Reasoning models finish their answer. Some setups returned the model's thinking but no reply at all; Mem now recognizes that and retries with room to finish.
Importing a folder stays responsive while it works, and the app keeps going when the underlying graph reports a storage problem instead of quietly retrying broken work in the background.
Browser extension: the Web clip toggle applies to the tab you are on, answers from Google AI Mode come across whole, and toggle labels respond to a click.

0.10.54

Aug 7, 2026

0.10.54Aug 7, 2026

Your Vocabulary Starts Doing The Work

Added

Naming the kinds of things you care about now changes what Mem writes down, not just how it shows you what it already has. Turn on the extraction switch in the Types tab and Mem reads new material looking for the vocabulary you defined. It is off by default, and while it is off nothing about your graph changes.
The Types workbench speaks your language. It arrived in English and Simplified Chinese only; it now reads natively in Traditional Chinese, Japanese, Korean, Spanish, French, German and Portuguese as well.
The browser extension shows you the article it is about to save, on the page itself. You decide next to the thing you are deciding about, instead of in a panel off to the side. Saving can also carry labels and a note now, folded away until you want them.

Improved

Large libraries scroll smoothly. Opening a source keeps its details and actions within reach instead of pushing them off screen.
AI Now picks up new models as they appear, lets you set how hard a model should think, and can be told to stop asking for approval on every step.
The one-time local license is called Mem Vault everywhere, in every language. Licenses you already own are unaffected.

Fixed

One damaged PDF no longer takes the whole import down with it. Mem reports that single document as failed and carries on with the rest.
The Types diagram stays centred when you resize the window. It used to drift off to one side and stay there.
Browser capture fixes: saving a PDF works on Firefox, conversations from X in languages other than English sync properly, and clip previews are readable again.
Scrolling a long conversation no longer loses your place, the Skills gallery uses the full width of a wide window, and Memory Health catch-up reads more clearly.

0.10.53

Aug 5, 2026

0.10.53Aug 5, 2026

The Shape Of What You Know

Added

The Knowledge Graph has a Types tab. It shows the kinds of things your knowledge is actually made of, and lets you name them, group them, and see how they connect as a diagram. Open it and the canvas becomes a lens on those types; leave and your explore view is exactly as you left it. If you work across languages, one type can carry its names in each of them, so a colleague searching in Chinese and one searching in English land on the same thing.
A conversation can now be handed to another tool by reference. Copy a thread and you get a link the AI can follow back to the full conversation in Mem, instead of a wall of pasted text.

Fixed

Refresh Search no longer spins forever. When a remote provider stopped responding mid-rebuild, the progress indicator had no way to give up and sat there indefinitely. It now reports that the rebuild stalled, and it will not start a second one on top of a worker that has not finished.
Reading text out of images uses a model that can actually see them. Some managed-AI setups from earlier versions were quietly sending pictures to a text-only model, which cannot do the job.
Restoring from a backup brings your AI Now conversations back too. They live outside the knowledge graph, so they used to be left behind when you moved to a new machine. Device paths and credentials stay out of the backup.
Grok models selected from a subscription connect on the first try, and the Library's add menu stays usable while an import is running.
Search is quicker on large libraries, and a run of smaller fixes from recent reports: clearer messages when text extraction has trouble, and steadier behaviour across the app.

0.10.52

Aug 4, 2026

0.10.52Aug 4, 2026

Save The Page, Not Just The Chat

Added

The browser extension now clips ordinary web pages and PDFs, not only AI conversations. Read something worth keeping, save it, and it lands in your Library with the rest of your knowledge. Captures happen on your own machine, and you get a preview of what will be saved before it is.

Fixed

A conversation captured from the browser keeps its middle. Turns could go missing from the middle of a long chat while the beginning and end came across fine, which is the kind of gap you only notice much later, when the reasoning no longer follows.
Answers from reasoning models are no longer cut short. They get the full length you set, and a reply that runs long now has room to finish.
Updating a Memory from an AI tool no longer clears its labels when you were only changing something else.
Restoring settings from a backup works when that backup includes your Spaces.
The Raft connection now lives with the other Connectors instead of off on its own, and the Timeline fills a wide window instead of leaving a column of empty space.

0.10.50

Aug 2, 2026

0.10.50Aug 2, 2026

Straight From Your Reports

Fixed

Connecting Grok through a SuperGrok subscription works again. Mem was sending the model name with a `supergrok/` prefix in front of it, and the service answered that no such model exists, so a perfectly good subscription looked broken. Mem now sends the plain name, and new connections default to Grok 4.5.
Hermes finds Mem again on macOS. A desktop host can start Hermes with a trimmed search path that leaves out the usual places a command line tool installs to, and Hermes would conclude Mem was not there at all. It now looks in those places before giving up.
The Kimi Code plugin no longer ships a hard-coded connection to a local Mem. Run `nmem config mcp show --host kimi-code` and it writes the server block for whichever Mem you actually use, local or remote, so the address and key have one source of truth instead of two that can disagree.
Reading a PDF is faster and cheaper. A page whose only images are small decorative marks no longer triggers a vision pass, so a long text document stops paying for one on every page.
When text extraction on a Library file goes sideways, the notice tells you what actually happened instead of a generic hint. A provider whose only job is embeddings is now labelled that way in Settings, rather than asking you for a default chat model it will never use.
Importing a session from the command line accepts a client name with stray spaces around it, which is how some hosts pass it.

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

0.10.62

Aug 13, 2026

0.10 is a ground-up engine rewrite

0.10.62

Aug 13, 2026

0.10.62Latest releaseAug 13, 2026

DeepSeek Harness, ZCode, and Safer Backups

Highlights

DeepSeek Harness and ZCode are now first-class Mem integrations.
Backups restore Library original files, not just parsed text.
Long remote-embedding rebuilds are bounded before they reach your provider.

Added

Fixed

Cloud restore now brings Library originals back with the backup. PDFs, documents, images, and other uploaded files can be previewed or downloaded after restore instead of leaving only parsed text behind.
Search Index rebuilds no longer send oversized Memory or Library text straight to a remote embedding provider. Long items are bounded before the request, so one large note is much less likely to keep semantic search stuck in text-only mode.
ChatGPT custom MCP connector setup is more reliable when OAuth discovery is probed after Allow. The pairing page now answers the compatibility endpoint without widening the connector's default read-only scope.
Data Transfer and Library recovery now fail more explicitly: recoverable export and restore mismatches become warnings, while binary-looking Library uploads are quarantined instead of being indexed as broken placeholder text.
AI Now no longer automatically replays slow or interrupted model requests. If a local or self-hosted model is still working, Mem now stops and lets you continue deliberately instead of queuing the same large prompt again.
The Threads source filter now treats case variants as the same source. Sessions from Hermes, hermes, and related Hermes agent names appear under one clean source instead of duplicate filters.
Desktop installers and self-hosted Docker images are now built with a more resilient PDF parser dependency download, so brief network interruptions during release packaging are less likely to delay an update.

cloud-2026-08-13

Aug 13, 2026

cloud-2026-08-13Aug 13, 2026

A Better Home for Teams on Nowledge Cloud

A wave of improvements for team workspaces on Nowledge Cloud. They arrive with your workspace, so there is nothing to install or update.

Highlights

Every Mem you belong to is one switch away, with Personal and Team clearly told apart.
Team imports explain who can read what before you commit.

Added

See every Mem you belong to and switch between them. Workspace pickers now tell Personal from Team at a glance, so you always know whose memory you are looking at.
Name your workspace. Rename it right from the switcher, and a new team is invited to pick its name the first time it opens.
Plan and seats in one place. Your team's plan, seat usage, and a direct path to billing now live together on the Team page.

Improved

Signing in takes you to your team first, so shared work is one step closer.
Application and sign-in emails now arrive in your language and address your company by name.
Invites tell the truth. If an invite email could not be delivered, Mem says so clearly, instead of leaving a teammate waiting on a message that never comes.
Importing into a team workspace now explains up front who will be able to read what, before anything is brought in.

0.10.59

Aug 13, 2026

0.10.59Aug 13, 2026

Clearer Setup and Graph Feedback

Highlights

Graph Explorer no longer gets stuck on an empty load.
Distillation tells you when a language model is missing.
Update install stays explicitly user-confirmed.

Fixed

Graph Explorer inside MCP-capable clients no longer sits on an empty loading screen. If the host opens the graph before sending data, Mem now tells you the next action to run.
Graph Intelligence is more reliable when you ask about the selected item. Mem now keeps the selected graph object in the chat context even when the canvas sends it through a wrapped event shape.
Conversation distillation now explains the real fix when a language model is missing. Downloading the Search Index model is not enough for distillation; Mem points you to the Default or Agents language model setting instead of showing a raw 422 error.
The update action is clearer after a download finishes. The title-bar update button now asks for the same explicit install confirmation as Settings, and Mem still never installs or restarts from a background download without your action.

Improved

Vision settings now say what actually happens in every language: if you do not choose a remote vision provider, Library uses the on-device local vision model rather than inheriting your default text model.
Long-running Codex and AI Now search cards stay compact while they are still working, then open when results are ready, so a single tool call no longer takes over the whole conversation.
Busy graph canvases get a quieter ontology legend that stays compact until you expand it or interact with a type.

0.10.57

Aug 12, 2026

0.10.57Aug 12, 2026

Connector and Library Reliability

Highlights

ChatGPT connector setup works through Access Anywhere again.
Oversized Library documents are bounded before embedding.
Kimi Code gets a smaller connector schema.

Fixed

ChatGPT custom MCP connector setup can complete through Access Anywhere again. After you click Allow, Mem now lets the authenticated setup flow read the status and pairing endpoints it needs, while still blocking public management access.
Library indexing is stricter about bad or oversized documents. Very long markdown, docx-derived text, and unusual source content are split and bounded before embedding, so one broken item should not take the whole refresh down or overload your provider's context limit.
All Spaces in Library now opens a source with the source's real Space behind it, so parsed content and details do not disappear just because you were browsing the aggregate view.
Windows launch-at-login is cleaned up at startup when the setting is off, so an old OS auto-start registration no longer brings Mem back after you disabled it.
Backups can now continue when an old Thread is missing historical message payloads. Mem records the missing content as a warning in the export manifest instead of blocking the rest of your data from being saved.

Improved

Kimi Code now gets the compact read-only connector profile instead of the full app tool surface, keeping setup lighter and tool schemas smaller.
Update downloads avoid cellular connections even when background downloading is otherwise allowed. Mem still downloads only in the background and never installs or restarts without your action.

0.10.56

Aug 11, 2026

0.10.56Aug 11, 2026

Memory That Keeps Growing

Fixed

Mem keeps learning as you use it more. Some of you told us your memory count stopped moving for days, in one case about a week, while you were working hard every day and everything looked switched on. Background learning had drifted apart from the part that actually writes a memory down, so the app could look busy without learning anything. It is one path again, and it is checked.
Cleaning up is something you decide, not something that happens to you. Maintenance now shows you what it wants to review before anything moves, so a number on a card is no longer a thing you have to trust blindly. Nothing is thrown away quietly.
Wiki topics refresh again, and their summaries follow your newer reading instead of stopping at what you knew months ago. The graph is quieter too: fewer orphan and near-duplicate entities cluttering what you are trying to read.
Search puts trustworthy memories first and does it before deciding which page you are on, so a good result can no longer fall off the end. Date filters on Memories work as you would expect.
AI Now is steadier: tools with unusual inputs no longer get refused, moving between tasks keeps its place, an idle session says it is idle, and a subagent uses the model you picked for the session.
A backup that could not capture a conversation completely is refused instead of written, so a restore does not hand you a half-finished thread.

Added

Share to Mem from your phone. Anything you can share on iOS can go straight into your Library without opening the app first.
Cindy and Mirasim join the connectors, and OpenCode works on Windows.

0.10.55

Aug 9, 2026

0.10.55Aug 9, 2026

Every Agent You Already Use

Added

Session sync now finds conversations from every AI agent Mem knows how to read, not just the five it used to look for. If you work in Hermes, Gemini CLI, Kimi, Grok, Antigravity or Pi, those conversations are waiting when you open session sync, and you choose which are worth keeping.
Mem now ships a portable plugin for tools that follow the Agent Plugins standard but do not have a dedicated Nowledge connector yet. It gives that tool search, memory saving, and handoff summaries. Where a dedicated connector exists, keep using it: it does more.
Amp now has a native Nowledge Mem connector. New Amp sessions can start with your Context Bundle, use memory and graph tools directly, and save real Amp threads automatically after agent turns.
Google Antigravity now has a dedicated Mem plugin, thanks to abn's contribution. Antigravity sessions can start with your current context, use Mem MCP tools and skills directly, and save real local transcripts automatically when a session ends.
Ekko Agent in Hermes Studio can now bring its exact local conversations into Mem. Ekko still uses MCP for context and memory tools; the new `nmem t sync --from ekko` importer reads completed runs from Ekko's SQLite message store and saves them as searchable `source=ekko` Threads.
Mirasim joins the connector guide. Add Mem as a Mirasim MCP server for the built-in agent, bring in the shared skills when your build supports them, and keep Claude Code or Codex panes on their dedicated connectors for exact thread capture.
Cindy is now in the connector guide too. Cindy's built-in agent can use Mem through MCP and shared skills; sessions that Cindy runs through Codex, Claude Code, Pi, or OMP keep their dedicated connectors for exact transcript capture.

Fixed

Refreshing search finishes, or tells you it failed, instead of running all night. A rebuild leaning on a remote provider that stopped answering could keep working for hours with nothing to show; it now gives up on a clear deadline. A rebuild wrongly reported as stalled can recover and finish rather than staying stuck on screen.
Reasoning models finish their answer. Some setups returned the model's thinking but no reply at all; Mem now recognizes that and retries with room to finish.
Importing a folder stays responsive while it works, and the app keeps going when the underlying graph reports a storage problem instead of quietly retrying broken work in the background.
Browser extension: the Web clip toggle applies to the tab you are on, answers from Google AI Mode come across whole, and toggle labels respond to a click.

0.10.54

Aug 7, 2026

0.10.54Aug 7, 2026

Your Vocabulary Starts Doing The Work

Added

Naming the kinds of things you care about now changes what Mem writes down, not just how it shows you what it already has. Turn on the extraction switch in the Types tab and Mem reads new material looking for the vocabulary you defined. It is off by default, and while it is off nothing about your graph changes.
The Types workbench speaks your language. It arrived in English and Simplified Chinese only; it now reads natively in Traditional Chinese, Japanese, Korean, Spanish, French, German and Portuguese as well.
The browser extension shows you the article it is about to save, on the page itself. You decide next to the thing you are deciding about, instead of in a panel off to the side. Saving can also carry labels and a note now, folded away until you want them.

Improved

Large libraries scroll smoothly. Opening a source keeps its details and actions within reach instead of pushing them off screen.
AI Now picks up new models as they appear, lets you set how hard a model should think, and can be told to stop asking for approval on every step.
The one-time local license is called Mem Vault everywhere, in every language. Licenses you already own are unaffected.

Fixed

One damaged PDF no longer takes the whole import down with it. Mem reports that single document as failed and carries on with the rest.
The Types diagram stays centred when you resize the window. It used to drift off to one side and stay there.
Browser capture fixes: saving a PDF works on Firefox, conversations from X in languages other than English sync properly, and clip previews are readable again.
Scrolling a long conversation no longer loses your place, the Skills gallery uses the full width of a wide window, and Memory Health catch-up reads more clearly.

0.10.53

Aug 5, 2026

0.10.53Aug 5, 2026

The Shape Of What You Know

Added

The Knowledge Graph has a Types tab. It shows the kinds of things your knowledge is actually made of, and lets you name them, group them, and see how they connect as a diagram. Open it and the canvas becomes a lens on those types; leave and your explore view is exactly as you left it. If you work across languages, one type can carry its names in each of them, so a colleague searching in Chinese and one searching in English land on the same thing.
A conversation can now be handed to another tool by reference. Copy a thread and you get a link the AI can follow back to the full conversation in Mem, instead of a wall of pasted text.

Fixed

Refresh Search no longer spins forever. When a remote provider stopped responding mid-rebuild, the progress indicator had no way to give up and sat there indefinitely. It now reports that the rebuild stalled, and it will not start a second one on top of a worker that has not finished.
Reading text out of images uses a model that can actually see them. Some managed-AI setups from earlier versions were quietly sending pictures to a text-only model, which cannot do the job.
Restoring from a backup brings your AI Now conversations back too. They live outside the knowledge graph, so they used to be left behind when you moved to a new machine. Device paths and credentials stay out of the backup.
Grok models selected from a subscription connect on the first try, and the Library's add menu stays usable while an import is running.
Search is quicker on large libraries, and a run of smaller fixes from recent reports: clearer messages when text extraction has trouble, and steadier behaviour across the app.

0.10.52

Aug 4, 2026

0.10.52Aug 4, 2026

Save The Page, Not Just The Chat

Added

The browser extension now clips ordinary web pages and PDFs, not only AI conversations. Read something worth keeping, save it, and it lands in your Library with the rest of your knowledge. Captures happen on your own machine, and you get a preview of what will be saved before it is.

Fixed

A conversation captured from the browser keeps its middle. Turns could go missing from the middle of a long chat while the beginning and end came across fine, which is the kind of gap you only notice much later, when the reasoning no longer follows.
Answers from reasoning models are no longer cut short. They get the full length you set, and a reply that runs long now has room to finish.
Updating a Memory from an AI tool no longer clears its labels when you were only changing something else.
Restoring settings from a backup works when that backup includes your Spaces.
The Raft connection now lives with the other Connectors instead of off on its own, and the Timeline fills a wide window instead of leaving a column of empty space.

0.10.50

Aug 2, 2026

0.10.50Aug 2, 2026

Straight From Your Reports

Fixed

Connecting Grok through a SuperGrok subscription works again. Mem was sending the model name with a `supergrok/` prefix in front of it, and the service answered that no such model exists, so a perfectly good subscription looked broken. Mem now sends the plain name, and new connections default to Grok 4.5.
Hermes finds Mem again on macOS. A desktop host can start Hermes with a trimmed search path that leaves out the usual places a command line tool installs to, and Hermes would conclude Mem was not there at all. It now looks in those places before giving up.
The Kimi Code plugin no longer ships a hard-coded connection to a local Mem. Run `nmem config mcp show --host kimi-code` and it writes the server block for whichever Mem you actually use, local or remote, so the address and key have one source of truth instead of two that can disagree.
Reading a PDF is faster and cheaper. A page whose only images are small decorative marks no longer triggers a vision pass, so a long text document stops paying for one on every page.
When text extraction on a Library file goes sideways, the notice tells you what actually happened instead of a generic hint. A provider whose only job is embeddings is now labelled that way in Settings, rather than asking you for a default chat model it will never use.
Importing a session from the command line accepts a client name with stray spaces around it, which is how some hosts pass it.

Ready to try Mem?

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

Download Read Docs