security
API keys can no longer create organizations
Creating an organization now requires a signed-in user. Requests made with an API key get a 403 error.
Changelog
What shipped in Memsy — new capabilities, improvements, fixes and security updates, newest first.
Subscribe via RSSOctober 2026
added
Sign in with AgentID
AI agents with an AgentMail inbox can sign in to the Memsy console with AgentID. Each agent gets its own account, and its owner is recorded.
fixed
Related memories keep all their details
When Memsy grouped related memories, some details on the group could be lost. New groups now keep every field.
fixed
Signing in no longer shows an error page
Some users saw an "Application error" page right after signing in to the console. That's fixed.
added
A customizable Home dashboard in the console
The console now opens on a Home dashboard with widgets for your memory insights and a Studio drop-in.
added
Filter memories, events and search by source
Every event and memory now records where it came from, such as a connector, Studio or the API. Filter by source on the Memories, Events and Search pages, or pass a source filter in search requests.
added
Invite a teammate on the free plan
Free tiers now include one member seat. Admins can invite a teammate and try working together before upgrading
September 2026
added
Cookbooks in the docs
Ready-to-run recipes for a Next.js chat app, a FastAPI support agent, a Slack bot, and memory as an agent tool.
added
Claude Code can search just this conversation
The Claude Code plugin can limit a search to the current conversation or skip it, and files each turn under the right conversation, including after /clear.
added
Memsy Studio — playground for memory creation
Type or paste text, drop in a documents, and watch it become memory. No SDK and no API key.
fixed
Connecting a source is more reliable
Cancelling at the provider's sign-in page returns you to the console instead of an error, and retrying Connect no longer ends in "Connector not found". The API also stays available during large document syncs.
fixed
Connectors created via API or SDK now sync
Connectors set up with an API key or the SDK showed as active but never ingested anything. They now sync like ones created in the console.
improved
More accurate search ranking
Closest matches now rank first. A scoring bug had pushed exact matches down, and broad sub-queries could outrank your actual question.
added
Memories stay in the language you wrote them in
Memories extracted from non-English content are now stored in the original language instead of being translated. Mixed-language content is handled as written.
added
Sign in to the S3 connector with SSO
The S3 connector now supports AWS SSO sign-in along with the pre-existing access keys login.
added
See who an event came from in your connectors
Events from Slack, GitHub, Google Drive, OneDrive, Notion, and S3 buckets connected with SSO now record the author's email in their metadata, so you can trace a memory back to the person whose content produced it.
improved
Scope a search to one conversation or exclude one
Search can now be limited to a single conversation, or told to skip one, including documents from your connectors. Send neither and search covers everything, as before. Available in the Python and Node SDKs (0.3.5).
fixed
Search finds all your memories again
A background job had cleared the search data for some memories, so search could miss them. This is fixed, affected memories are restored, and new memories keep their search data from the moment they're saved.
improved
No more filler memories
Replies like "Thanks." or "Got it." are no longer saved as memories, and memories keep their links to the events they came from.
fixed
Score thresholds work again
Relevance scores now stay between 0 and 1, so the threshold option filters results as documented.
added
Browse actors across your organization
New Actors tab on the Hierarchy page lists every actor in your organization with their events and memories. Also available in the Python and Node SDKs (0.3.5) and the MCP server.
August 2026
fixed
Console loads for large organizations
Memory and stats queries in the console are now paginated, so organizations with large memory counts load instead of timing out.
improved
Search and indexing retry transient failures
Reads and idempotent writes to the vector store now retry on transient faults, so a brief backend blip no longer surfaces as a failed request.
fixedsecurity
Retrieval no longer crosses session boundaries
Context-window lookups now require a session id. Previously a request without one could return results drawn from other sessions in the same organization.
added
Promotion visualization on every plan
The promotion graph, which shows how individual memories were consolidated, is now available on the Free tier rather than Pro and above.
July 2026
improved
Faster reranking, now on Pro and above
Reranking has a latency cap, so a slow provider can no longer stall a search, and is enabled for Pro, Growth and Enterprise plans.
fixed
Sign-in works on the custom domain
Content security policy now allows the authentication scripts served from our own domain, fixing a blank sign-in screen for some users.
fixed
Connectors visible by default
The connectors section is now shown unless explicitly disabled. Previously a deployment that did not set the flag hid it behind a "coming soon" badge with no indication anything was wrong.
fixed
Organization admins can no longer be locked out
An admin without an assigned seat was refused access to their own organization. Seat assignment is no longer required to administer it.
fixed
Sign-up CAPTCHA loads correctly
The bot-protection challenge on sign-up was blocked by content security policy, which could prevent account creation entirely.
added
GitHub connector
Connect a GitHub organization and sync repository content into memory.
added
OneDrive connector
Connect OneDrive and pick specific folders or files to sync, rather than syncing an entire drive.
added
Notion connector
Connect a Notion workspace and sync pages into memory.
added
Amazon S3 connector
Connect an S3 bucket, scoped to your organization, and sync its documents into memory.
June 2026
added
Google Drive connector
Connect Google Drive and sync documents into memory.
added
Promotion insights
See how memories were clustered and promoted for your organization, with per-cluster detail.
added
Slack connector
Connect one Slack workspace per organization. Connection status is visible to members, not just admins.