Strata PlatformDocs

Personal settings

Settings is where you configure Strata for yourself — your profile, appearance, memory, privacy, and the apps you run Strata through. Open it from the account menu at the bottom of the sidebar; it opens as an overlay on top of whatever you were looking at, so changing a preference never costs you your place.

The overlay has a side navigation down its left: Search settings at the top, then two named groups, Settings (General, Account, Privacy, Capabilities, Memory, Usage and Recap) and Customize (Apps), with the section you chose beside them. Every section is deep-linkable — /settings/general, /settings/privacy, and so on — so you can bookmark or share a link straight to one.

On a phone, Settings opens on its list of sections, in the same groups; choosing one opens it, with ‹ Settings above it to go back to the list (your browser's Back does the same). A link to a section opens that section directly.

To find a setting by name, type in Search settings at the top of the side navigation. Pick a result with the mouse, or with the arrow keys and Enter, and Strata opens its section and briefly highlights the setting; Esc clears the search. Organization has the same search for administrators.

Note

When your administrator has pinned a setting organization-wide, the control carries a small padlock badge reading Managed by your organization's name, and stops responding. You see the current value but cannot change it. Hovering the control repeats the reason.

General

The General section covers your identity, profile, personalization, and how the app looks on this machine.

Identity

The section opens on you: your avatar at full size, your display name, and one line showing where you sit — email, organization, and role. Your profile is signed in through Microsoft Entra, noted as Managed by Microsoft Entra.

  • Change photo — opens the avatar picker, where you choose your initials or one of ten abstract symbols. What you pick is your picture everywhere in Strata: the sidebar, the admin pages, announcements you send and Strata for Outlook. Hidden when your admin has locked the avatar.
  • Edit name — jumps to the Full Name field below. Hidden when your admin has locked the display name.

Profile

  • Full Name — synced from your Microsoft account by default and editable in place unless your admin has locked it. When your organization turns off Let members change their display name, the field shows your Microsoft name, or a name an admin set for you, which that switch doesn't replace.
  • What should Strata call you? — a short preferred name or nickname (up to 50 characters) Strata uses when addressing you.
  • What best describes your work style? — a single-select with four options: Analyst (data-focused, prefers charts and metrics), Technical (shows the working, prefers code and detail), Executive (high-level summaries and insights), and Operations (process-oriented, step-by-step). Strata uses your work style to bias how it phrases answers.
  • Custom Instructions — a free-text box (500-character limit) where you describe your preferences, role, or how you'd like responses formatted. Applies to every conversation.

Personalization

Two rows, each opening a modal, control how Strata writes as you:

  • Writing voice — Strata matches how you already write. You can let it learn your voice from your own messages.
  • Signature — the sign-off Strata adds to emails it drafts for you. Strata can find it at the end of your sent mail, or you can type or edit it yourself.

Language

Language sets the language of Strata's menus, buttons and messages. It sits at the top of the card with Timezone and Appearance and shows the language you use now, in its own name. Change opens Choose your language, which you can also open from Language in the account menu.

The dialog lists 35 languages, each in its own name over its English name, with yours checked. Type in the search box to narrow the list by either name; accents do not matter. Your choice applies at once and is saved to your account, so it follows you to every device you sign in on. Until you choose, Strata uses your browser's language when it is one of the 35, and English otherwise.

Strata replies in the language you write in, and asking for a language always works. When a message has no words of your own to go by (a file, a link or pasted text sent on its own, or a reply such as "ok"), it replies in the language you use Strata in, and when a new chat's first message does not settle it, the chat's name is in the language you chose here. In Teams, if you have not chosen a language, Strata goes by your Teams language. Something Strata writes for other people, such as an email, follows its readers rather than your setting.

The languages are English (US), Amharic, Bangla, Bhojpuri, Chinese (Simplified), Chinese (Traditional), Danish, Dutch, Filipino, Finnish, French (France), German (Germany), Gujarati, Hausa, Hindi (India), Indonesian, Italian, Japanese, Kannada, Korean, Malayalam, Marathi, Norwegian Bokmål, Portuguese (Brazil), Russian, Spanish (Latin America), Spanish (Spain), Swahili, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian and Vietnamese.

So far the sidebar (with Edit sidebar), Settings, the chat (its composer and menus, the greeting and suggestions of a new chat, the buttons and notes around each answer, the one-line summary of the steps Strata takes while it works and how long they took, and the cards it draws inside an answer: query results, charts, diagrams, tables, comparisons, timelines, maps and places, the calendar, the Jira cards, email drafts, generated files and their downloads, the questions and approvals it asks you for, requests to connect an app, and the workspace with its checklist, outputs, sources and sub-agents; voice mode and its messages, dictation, attachments, mentions, quoting, the connection notice and a chat exported as a PDF too, with times, durations and numbers written the way your language writes them), the Customize page (its tabs, search, filters, cards, and the request Try opens in a new chat), All chats (its search, day groups, selection and the delete confirmation), the Inbox (its list, filters, search and empty states, the welcome message, the SQL Server disconnected notice, and the words around every message: who sent it, who it went to, its priority, Ask Strata about this and when it leaves your inbox), the Data pages (Tables, Procedures, Explorer and Schema map: their tabs, search, filters, sorting, counts and messages, with numbers and dates written the way your language writes them), the Agents page (the list and its filters, each agent's page, its runs and a run's details, the form for creating and editing an agent, the templates, the webhook trigger, and every message and confirmation, with schedules, times and dates written the way your language writes them) and the Agent draft card Strata shows in a chat when it drafts an agent for you are translated, as is the setup wizard a new organization's first admin walks through (Onboarding your organization: its seven steps, the trial and seat sentences and every answer of the Foundry check), and the Directory that Apps opens: its Skills, Extensions, Connectors and Procedures tabs, a skill's page, every extension's page with what your organization decided about it and, for an admin, its Manage page (the Atlassian site, the connection style and the Jira and Confluence switches among them), your saved SQL Server connections with the form that adds one, and a connector's page, and the editors a file's preview opens with Edit (its pencil button) for a Markdown file, a Word document, a PDF report, a deck and a workbook (their tools and menus, the colour choices, the line that says what changed and the question before your changes are discarded), with the spreadsheet preview and its notices; other pages are in English for now. Numbers on those pages are written the way your language writes them, including the counts beside the sidebar's sections and lists, file sizes and the numbers on charts; Bengali and Marathi use their own digits. Dates, times, how long ago something happened, lists of names and units are written your language's way in every browser; Chrome and Edge have no Hausa of their own, so Strata brings Hausa's with the language. In the chat, the names of the steps and the line Strata shows while it works (such as Ran query, Searched the web, Searched Jira or Ran 3 sub-agents) are in your language too, in chats from before you changed it as well, including the steps it takes when it runs a stored procedure, creates, reviews or edits a file, writes a document or deck part by part (such as Wrote 6 sections · 2,310 words, with the line under it that says how far the writing has got), runs Python or sub-agents, looks up places, finds earlier chats or knowledge-base passages, works in Jira, Confluence, Outlook, OneDrive, SharePoint or Teams, updates the checklist (such as Checked off 2 items) or waits to retry, and the line that says an answer stopped because a credit allowance ran out or a model is turned off or out of allowance; the line under a step that failed (such as Jira is not connected) is in your language too, except words that come from elsewhere: a database's or another service's own error message is shown as it wrote it, and a note Strata writes for the AI model, such as why an approval did not go through, stays in English; Jira's own words, such as a status or a priority, appear as your Jira site writes them; and if the full PDF export is unavailable, the plain-text copy Strata falls back to is in English. Words written once and kept stay in the language they were written in: an announcement's title and text, an agent run's summary, and the notices Strata posts from its servers (a monthly credit allowance reached, a model retiring). What an editor adds to a file starts in your language: a new section's heading, a new table's columns, new key figures, a callout's label, a new text box, a new sheet's and a new column's names (for example Tabelle3 and Spalte 5 in German) and the columns of a table in Markdown; the words already in the file stay as they were written, and a workbook's cells show their numbers as the workbook's own formats write them. The names and descriptions of skills and connectors stay as they were written, as do product names, model names and keyboard shortcuts; the exception is a connector from Strata's catalog, whose description is in your language. The groups they are sorted into (such as Create and Analyze for skills, or a connector's category) are named in your language, and so are Web search and Teams chats, which are named as the Organization page names them. On the Data pages your database's own names (its schemas, tables, columns and procedures), its data types and SQL words such as NOT NULL and NULL stay as they are, since they are what you would type in a query. An agent's name, description and instructions stay as they were written, whether by you or by Strata in the chat that drafted it, and so do your organization's own agent templates; the built-in templates are translated, including the instructions they fill in. A run's steps and error messages are recorded in English as the run happens and are shown as recorded. On the Organization page the frame is in your language: the list of sections and its search (a setting in a section that is still in English is found by its English name), the title and description of each settings section, the notes that say a section is loading, could not load, is waiting for your permissions or needs a permission you do not have, the banner shown while the organization is suspended, and Show more under a long description. So is all of Profile: its settings, the export, suspend and deletion dialogs, and the notice that shows a requested or scheduled deletion. The number fields in Member defaults, Agents, Profile and Data and privacy take a number written the way your language writes it (for example 1.234,5 in German) as well as the English way, and Member defaults and Agents show their larger numbers that way too. So is all of Agents: its limits and the notes under them, the list of your organization's templates, a template's page and who is using it, the template form, and their menus, confirmations and messages, with each count said the way your language says it; your organization's own templates keep the names, descriptions and instructions they were written with. So is all of Support: the note before you write and the email line, your tickets and their filters, each ticket's line (its category, its messages, who replied last and when), the notice that Kronisys has replied, the New ticket form and a ticket's conversation, and their confirmations and messages, with counts and times written the way your language writes them; a ticket's subject and the messages in it stay as they were written. So is all of Member defaults: its settings and their descriptions, the choices in its menus and the words beside its padlocks, the autonomy limits and the limits this deployment sets, the line under a capability that says what your organization, its roles and its members have set, and its confirmations and messages, with each count said the way your language says it. So is all of Roles: the list with its search and count, a role's view and the role editor in its four tabs, the delete confirmation, and their messages, with each count said the way your language says it. The two built-in roles are named and described in your language across the Organization page, its Users section and role menus included (Admin and Member in English); a role your organization made keeps the name and description it was given. So is all of Extensions: the choice between Ceiling and Menu and what each means, every extension's description and the line that says how many roles list it, the warning shown before an extension goes off or Menu becomes Ceiling (who loses it, who is disconnected and what turning it back on does), the line shown while the change is saved, the notes for an extension this deployment has not set up or one that needs another to be on, and the closing line that says where connectors, roles and members are governed, with each count said the way your language says it; a role's Access tab describes the extensions in your language too. Web search and Teams chats are named in your language; product names such as Outlook and Jira stay as they are. So is all of Sessions: its search, sort and count, who is signed in, on which device and when they were last seen, a member's sessions with when each started and expires, and the confirmations for signing out one session, one member or everyone, with each count said the way your language says it; a device is named the way Settings names it (for example Chrome on Mac). So is all of Knowledge base: the SharePoint sites with their documents, passages, last sync and status, the summary and the per-site file limit, the notice that syncs depend on one admin's Microsoft account and the buttons that give the organization access, adding a site, and the confirmations for reindexing or removing one, with each count said the way your language says it; a site keeps the name it has in SharePoint. So is all of Groups: the list, a group's page with its activity figures, how they changed on the period before and the roles bound to it, the group's settings with its monthly spend limit, its model limits and its members, the picker that adds people, and the confirmations for deleting a group or removing someone from it, with every count, token figure, percentage and date written the way your language writes them. A model limit there can be typed the way your language writes numbers (for example 3.000.000 in German). A group keeps the name it was given, or the one your identity provider gives it. So is bringing people in and seeing them out: Add user (one person or several at once, each person's role and license, the notes on free licenses and seats, and what happened to each address), an invitation's row on Users and its menu (resend, configure before they join, copy the address, revoke), Pre-configure invitation, the note that your identity provider manages the roster, a member's license (Free or Billed), and the dialog that deletes a member, with what happens to the agents they own; a count, a date and a list of names are written the way your language writes them. So is the Users list itself: its search and group filter, each member's row (role, status, when they were last active), selecting people (everyone shown, or everyone who matches), and the bulk commands with their confirmations, progress and results: activate, disable, change role, add to or remove from a group, export and delete. Where the window is too narrow for all of the bulk commands (on a phone, for example), the bar keeps More and Delete, and More starts with Activate, Disable and Change role. A member's own window is in your language too: the summary at its top (their role, how many extensions and models they can use, when they were last active and when they joined), the Profile tab (their name and its lock, role, account status, license with what a free license does to your bill, rate limit and monthly spend limit, with amounts written the way your language writes money) and the Access tab (extensions, capabilities with what decided each one, the sub-agent approval limit, models and database permissions), with their confirmations and messages, and so are its other three tabs: Data & sessions (their groups, where each of their permissions comes from, signing them out everywhere, wiping their memory, deleting their conversations, the export that is emailed to you and what it says when it cannot be made, and deleting the account), Usage (the range, their turns, tokens, spend and reply time, the chart, the models they used and how often they reached their rate limit, with numbers and money written the way your language writes them) and Agents (what their agents may use, each agent's schedule, status and last run said the way the Agents pages say them, and deleting one). So is the organization's list of Database connections: each connection's health and when it was last tested, who has access to it, its server and database or the private connector it goes through, when it was created, its details, what testing, disabling, enabling and deleting one say, and the connection form with every check it makes before it saves. So are the Private connectors listed under it: what a private connector is for and how it works, each connector's agent and the connections that use it, the notes when one is still in use or when private connectors are not set up on your deployment, and New private connector, Setup and Remove with every check New private connector makes before it creates one; a reason that comes from Strata's servers, such as why a connector could not be created, is shown as the server wrote it, in English. A permission is named in your language wherever the page around it is, as on Roles and in the notes that say a section needs a permission; a page that is still in English names it in English (for example “Manage org settings”). On Usage, the top of the page is in your language: the window and its dates, the summary with how each figure moved on the window before, the heading, caption and controls of the chart and the table of models, Seats, Time saved and the note at the foot of the page, with amounts of money, counts and dates written the way your language writes them. So is the table of people: its heading, the switch between members and API accounts, the search, the columns, each row's role, status and when the account was last active, the count under it and the line about API accounts it holds back, and the column headings and words of both CSV exports. So are Key insights (each insight, and the menu of windows), the chart's own words (the names in its key, the dates under the bars, what the box shows when you point at a day) and the Compute credits card, down to the month's name and the time of each event. So is a member's or an API account's billing profile: its figures, its insights, its tables and their filters, an API key's settings, and its three CSV files. So are a member's conversations: the list and its selection, the export dialog, and a transcript with its files. So are the requests for a higher spend limit, on Usage and on Billing, and Spend limits with Edit limits: each request's figures and sentences, the two settings above the table, each group's limits and budget, the suggestions and the warning in Edit limits, and their confirmations and messages, with amounts of money, percentages and counts written the way your language writes them; a member's reason for asking stays as they wrote it, and a group keeps its name. So is the Audit log: the line above its table, the search, the categories, the date fields and their calendar, the table with its count and pages, the export dialog, and an event's details with the notes on its map, with dates and times written the way your language writes them, and a date can be typed the way your language writes it. What an event records stays as recorded: its name (such as user.role_changed), the people and addresses in it, the place its address was traced to and its details; the CSV export's column headings and categories stay in English. So is API keys: the switch for the organization's API access, the search, the filters, the table and its count, a key's menu, New API key with its owner list, scopes, calendar and messages, the key and the webhook secret shown once, a key's details (its scopes named as New API key names them, its rotation history, this cycle's usage and its webhooks), Edit settings, Rotate key, Revoke and the connection bots, with dates and numbers written the way your language writes them. A key's name and prefix, a bot's name, the event key.revoked and the Strata-Signature header stay as they are, and a message that comes from Strata's servers, such as why a key could not be created, is in English. So is Connectors: the table with its count, who can use each connector (its roles named and separated the way your language separates a list), its tools and its status, a connector's menu, Check all with the notes it leaves on each row and what it found, the catalog behind Add connector with its search and its seven categories, the connector's form (its address, sign-in or key, Test connection and the choice for each tool), and the confirmations for deleting a connector or leaving unsaved changes, with times and counts written the way your language writes them. A connector's name and address, its tools' names and descriptions and a message from its server or from Strata's servers (such as why a check failed) stay as they were written. The catalog's own words are in your language too: what each vendor's connector is for, where its credential comes from and what it may change; the vendor's menu paths and button names, addresses, scopes, tool names and a placeholder in angle brackets stay as written. So is Billing: the amount at its top and the note under it, its notices (a payment past due, more members than seats, a rate that does not match your size, a paused seat sync, a missing payment method, a trial), the account's rows (plan, seats with invitations and free licenses, contract, rate, billing period, collection, payment method, compute credits, credits and the response fee), the subscription's row with Cancel subscription and Keep my subscription and its confirmation, the invoices with their statuses, and the notes at its foot, with amounts of money, counts and dates written the way your language writes them (for example 1.515,70 $ in German). So is Manage billing: your plan, the spending limits, the default limit per person and the weekly pace with their notes, confirmations and messages, and the payment portal. So are Plans (the four plans, what each includes and the terms of the free trial) and the step where you choose how many seats, with its order summary; a plan keeps its name (Starter, Business, Enterprise and Global), and an invoice keeps the number Stripe gave it. So is Data and privacy: retention, the audit log's settings and the Tamper-evident log, Alerts for blocked requests, Member chat oversight, What can leave your organization and Conversations in feedback to Kronisys, with their confirmations and messages, and the top of Advanced data protection (its status, what it did lately, who changed it last, Reset breaker, Save data protection and Remove configuration… with its confirmation), with every count, date and number of days written the way your language writes them. Inside Advanced data protection, so are Who answers with the endpoint or vendor and its key, For your developers, What this control does not cover, Response inspection, Pilot and rollout, Test endpoint, Content detection with the detectors' names, Your own rules, Try it on some text with what it found, and the confirmations they ask for when you save, while the code samples stay exactly as your endpoint receives them. Each detector's own description and its Will get wrong note are in your language too, with the cue words, code and sample values they quote left as the detector reads them. So is Models: the summary of the Model Catalog, Models your members can use with its allowances and the dialog that changes them, Allowance defaults, Default model for new members with the notes under it, and Model retirements, with token amounts, retirement dates and counts written the way your language writes them; an allowance can be typed the way your language writes a number (2.000.000 in German) or as 20M or 500k in any language. Model and provider names stay as the catalog writes them. The Model Catalog that Manage models opens is in your language too: its list with the filters, search and sorting, each model's page (overview, capabilities, pricing, benchmarks and allowance), the model guide, the catalog's settings with the three steps that set up autonomous deployment, a deployment's Manage deployment page, and every confirmation and message its switches raise, the list of agents that would pause included, with prices, token figures, rates and dates written the way your language writes them and the two built-in roles named as the Roles page names them. A model's own description, strengths and what it is best for stay as the catalog writes them, as do a benchmark's suite and note, the names Azure gives regions, versions, deployment types and roles (such as GlobalStandard and Cognitive Services Contributor) and the command you copy to grant access; Azure's own menu paths, such as Access control (IAM), are written as the Azure portal writes them in your language, and in English where the portal does not come in your language. So is AI engine: where your organization's AI runs and whether Strata can reach it, Connect your own Foundry… (or Change AI engine…) with its endpoint, its key and the confirmation it asks for, the line that says how many of your enabled models are deployed on your Foundry, Deployment names for search, voice and transcription, and every message a save can give, with each count said the way your language says it; the endpoint, your deployment names and the names of models stay as written. So is Provisioning: the tenant URL your identity provider uses and its copy button, the tokens with who created each one, when it was last used and when it expires or was revoked, Create token with its expiry choices and the token shown once, the confirmation for revoking a token, and the recent changes your identity provider made, with dates, times and counts written the way your language writes them; a token's name, the people named in that activity, SCIM 2.0, Microsoft Entra and Okta stay as written. So is Announcements: the drafts, the Sent, Scheduled and Templates tabs with their filters and table, an announcement's details with who has read it, Read rates, the template dialog, and the page where you write one, with its ribbon and menus, the steps of Publish and the windows that insert a table, chart, callout, embed or image, with counts, percentages, dates and the numbers on a chart written the way your language writes them; an announcement's title and text, a template's name and description and the people and groups it goes to stay as written. A message from Strata's servers stays in English, such as why a save or a rule was refused or the marker that stands in for a masked value, and a tripped circuit breaker quotes the endpoint's own error as the server recorded it. The question Strata asks when you leave a page with unsaved changes is in your language; on a page that is still in English it names what you were editing in English. Choosing a language while the Organization page is open draws it again in that language, unless it holds a change you have not saved; then it keeps the language it was in until you next open it, so the change is not lost.

Timezone and Appearance

Timezone and Appearance share a card at the bottom of the section. Neither changes what Strata knows; both are saved to your account, so they follow you to every device you sign in on.

  • Timezone — used everywhere Strata shows timestamps, including chat history, dashboards, and activity reports. The top of the list offers Automatic, naming the zone your browser reports; below the separator is the full list of zones. May be locked by your admin.
  • Appearance — a three-way control: System (follow your OS preference), Light, or Dark. Strata is no longer dark-only; the whole app renders in your chosen theme, and System tracks your OS setting live.

The Strata version label sits quietly at the bottom of the section — read it out if you ever contact support.

Account

The Account section is your identity and the actions that cannot be undone.

  • Signed in as — the email on your account.
  • Organization — your organization's name and your role in it.
  • Organization ID — a copyable identifier. Support will ask for it if you raise a ticket.
  • Sign out everywhere — ends every active session, including this one, on all your devices. You'll need to sign in again.

Privacy

The Privacy section covers what Strata saves about your activity. An intro paragraph summarizes your controls and links to the full Privacy Policy and Terms of Service.

Note

If your organization has chat oversight turned on, a note at the top of this section says so: an administrator holding the right permission can open and export any conversation on this account, and every time one does it is written to your organization's audit log.

Data Privacy

  • Save AI responses to history — when on, your prompts and Strata's replies are saved so conversations reappear in history after you reload or come back. When off, nothing is kept — once you leave or refresh, the messages are gone for good. The same applies to your conversations with the support assistant in the Strata Help panel. Token usage is recorded either way.
  • Keep my extension-data chats in history — a separate control for any conversation turn that used a connected extension (SQL, files, email, knowledge base). You can only make this more private than your organization's setting, never less. When you turn it off, for any turn that touches an extension neither your message nor the reply is stored; the conversation still appears in your history, titled "Conversation", with a note in place of the text. Token usage is still recorded. When your organization sets this and locks it, the control is disabled.

Feedback

  • Help us improve Strata — when on, a thumbs up or down on an answer is recorded, and a short form opens asking what was good or what went wrong. On a thumbs down you can pick what went wrong (Wrong or made-up facts, Didn’t do what I asked, Incomplete, A step failed, Too slow, Looks broken, Something else). Send gives Kronisys, the makers of Strata, the rating, your note, and diagnostics of that answer: the model, how long it took, the token counts, and the tools it used with whether each one failed — never the words of the chat. Cancel keeps only the rating. The conversation goes too only if your organization allows it and you tick Include this conversation; deleting that chat later removes the copy. Kronisys keeps what you send for 90 days, and your organization’s audit log records that you sent feedback, never your note. When off, ratings are not recorded and the form does not open.

Export your data

An Export button starts the data export flow. See /docs/settings/data-export.

Active sessions

A table of the devices you've signed in from, showing the device, where it signed in from, and when it was last seen. The session you're currently using is marked Current. Use a row's menu to Sign out of any device you don't recognize; the current row has no menu — use Sign out everywhere in Account to end it.

The list shows one row per device, not one per sign-in. Repeat sign-ins from the same browser and network — after a password prompt, a token refresh, or simply revisiting later — are folded into a single row with a +N earlier sign-ins note. Signing out of that row ends all of them.

Location is resolved from the sign-in IP address and can take a moment to appear. Until it does — or for a private or corporate-VPN address that can't be placed — the row reads Unknown location followed by the raw IP, so you can still tell two devices apart.

Capabilities

The Capabilities section shows what Strata is able to do on your behalf. These reflect your organization's policy and your connected extensions; most are shown here for transparency rather than as personal toggles, and some are set by your organization and can't be changed here.

  • Open links you share — Strata can read a web page you paste and answer from what it says. Your organization can turn this off entirely or limit it to an allow-list of domains; private and internal addresses are never reachable.
  • Run code and build files — Strata can run Python against your data and produce spreadsheets, documents, PDFs, and decks. Code runs in an isolated sandbox with no network access.
  • Search the web — Strata can look up current information when a question needs it. Your organization controls whether this is available. When it has pinned web search with its padlock, the row carries a Managed by badge with your organization's name, shows your organization's setting, and reads Your organization has turned this on for everyone. (or off). The Web search row in the chat plus menu is locked the same way; see Web search.

Memory

The Memory section governs what Strata remembers across conversations.

  • Allow Strata to access memory — the master toggle. When off, Strata neither reads nor writes memory and starts fresh every chat. May be locked by your admin.

What Strata may use

Each of these is off until you turn it on, and each changes what Strata can actually do — not just what it is told:

  • Remember sensitive topics — lets Strata keep details about things like health, beliefs, politics, or background in long-term memory. With it off, Strata can still use what you tell it within a conversation but won't write those details down. Turning it off does not remove anything already saved.
  • Look things up in your past chats — lets Strata search your earlier conversations for anything relevant to what you're asking now, so you don't re-explain work you've already done. It only ever searches your own chats and tells you when it draws on one.
  • Use my approximate location — lets Strata know the city or region you're connecting from so times, dates, units, and "near me" questions come out right. City and region only — never a precise position, and your IP address is never given to the model.

Managing your memories

Your saved memories are listed on this page. Strata saves them automatically via a silent memory command; you can delete individual entries here but not edit them. A Bring your memory from another AI card with a Start import button opens a wizard for pasting exported memory text from another assistant.

See /docs/chat/memory for the full memory model.

Note

Procedure shortcuts (the stored procedures Strata has indexed, with per-procedure toggles for semantic routing) live in the SQL Server connection details, not here. Open them from the Directory's Procedures tab, which appears once a database is connected.

Usage

The Usage section shows your AI usage on Strata against your organization's limits.

  • Your AI usage — current consumption against your limit, applied on a rolling one-hour window.
  • Other limits that apply to you — any additional limits set specifically for you appear here; the card is empty when none are. A monthly usage limit shows the estimated model usage of your work this month against it, in dollars and as a percent, whether it was set for you, by a group you are in or by your organization, and that it resets on the 1st. When your organization sets a weekly pace, This week shows your usage since Monday against the week's share of your limit, and when it resets, in your own time. At 75%, 90% and 100% of your monthly limit, a banner over the message box says so too. When every shared budget you draw on has run out, a Team budget row says so, until the 1st or until an admin raises it. Request more usage under it asks your admins for more: add a note if you like. The request waits for an admin on the Usage page, and you are emailed the answer, which also appears in your inbox. The button is hidden while a request is waiting; once it is answered you can ask again whenever you need to.
  • Billable usage this period — what you personally have consumed this billing period. This is a charge, separate from the hourly rate limit above.

A refresh button updates the cards, and a Learn more about usage limits link opens the usage documentation.

See /docs/admin/rate-limits for how usage and limits are configured.

Recap

The Recap section is your own history with Strata, built from your conversations and visible only to you. There is no organization view here.

  • Period and refresh. The menu at the top right picks Past month (where Recap opens each time), Past 3 months, Past 6 months or Past year. The button beside it builds the recap again; hover it to see when the one on screen was made. Strata keeps a recap for up to six hours, so opening it again is quick. A short animation of a page being written plays while a recap is built (it stands still if your system asks for reduced motion).
  • The opening is a headline and a short summary of what you focused on, written by AI. When that part is not available it says how many conversations you had and how long you spent.
  • Your time with Strata shows your most active day, your peak hour, the conversations active in the period, and Total tokens, counted the way Settings › Usage counts them (with the part your agents used beneath). A line chart below shows the trend, by Conversations or Time spent.
  • What you spent time on is one bar split into up to five topics, each with its share of your time and a line about the work in it; the rest is Other. The topics are written by AI from the titles of your conversations in the period, and nothing else from them, in your organization's own Azure AI Foundry, on a model you are allowed to use; writing them never counts toward your usage limits. Topics about personnel matters concerning named people, health, legal matters or personal finances are never named: they count toward Other. If your chat history is not saved, no titles are sent and none are kept. The topics appear once you have had at least three conversations in the period, and deleting a conversation removes the topics written from it.
  • What Strata did for you counts Analyses, Files created, Database queries and Web lookups, each shown only when it is not zero.
  • Getting more from Strata offers up to three suggestions where your conversations support one: a prompt you can copy, an idea for a scheduled agent (when you can create agents), or something worth telling Strata to remember (when memory is on).
  • Your model is the model you used most, by conversation turns.

Apps

Under the Customize heading, the Apps section covers the ways to run Strata somewhere other than this browser tab, and the door to everything Strata can reach.

Strata for Outlook

The Strata task-pane add-in brings Strata into Outlook, docked next to whatever you're reading — same account, same models, same governance as the web app.

  • Set up in Outlook — a short guided setup that installs the add-in just for you.
  • Copy manifest URL — copies the add-in manifest URL for an org-wide deployment through Microsoft 365 integrated apps.

See /docs/integrations/outlook-add-in.

Directory

The Open Directory button opens the Directory — one place for everything Strata can reach: the skills it can load, the extensions you connect yourself (SQL Server, OneDrive, Outlook, Atlassian), and the connectors your organization has set up for the whole company. This is where you connect, disconnect, and manage each extension, and where you pin skills so Strata reaches for them without searching first.

  • Extensions — connect a service, view its connection details, and disconnect it. Each extension has its own page, deep-linkable at /settings/extensions/<id>, where What your organization has decided says whether it is available to you. If Strata could not check just then, it says Strata could not check whether <extension> is available to you just now. with a Check again button; an open page updates when the answer changes, after Check again or when you come back to the tab (Strata re-reads it then at most once a minute).
  • Skills — browse the skills library and pin the ones you use. Pinning a skill puts its name in front of Strata for every message in that conversation, so it doesn't have to go looking first; there is no limit on how many you pin. Strata still loads a skill's instructions at the moment it uses one — pinning shortens the finding, not the reading.
  • Connectors — the additional connectors an admin has enabled for your organization appear here alongside the built-in extensions.

See /docs/integrations for connection guides and /docs/chat/skills for the skills model.

Managed settings

Throughout the overlay, a setting your admin has pinned carries a small padlock badge beside its label reading Managed by your organization's name. The control still takes focus and still announces itself to a screen reader, but it does not respond — you see the current value and cannot change it, and hovering it says who decided. Ask your admin if you need one opened up: on their side it is a padlock on that setting's own row under Organization — or, for your display name, profile picture and timezone, a Let members… switch under Profile → Member profiles — so they can see what everyone is being held to before they open it.

Related