October 2026
Releases Shipped
Section titled “Releases Shipped”| Version | Released |
|---|---|
| v5.12.6 | 2026-10-01 |
| v5.12.7 | 2026-10-02 |
| v5.12.8 | 2026-10-03 |
| v5.12.9 | 2026-10-05 |
| v5.12.10 | 2026-10-06 |
| v5.12.11 | 2026-10-07 |
| v5.12.12 | 2026-10-20 |
| v5.12.13 | 2026-10-20 |
| v5.12.14 | 2026-10-21 |
| v5.12.15 | 2026-10-22 |
| v5.12.16 | 2026-10-28 |
Finance and Data Foundation
Section titled “Finance and Data Foundation”-
Data Foundation brings your ERP data together in one Common Finance Model. The Common Finance Model brings together ERP data from multiple ERP systems and instances into a single model: one chart of accounts, one set of entities and one reporting currency. Reports, metrics and AI assistants built on it all read the same numbers, whichever ERP they came from.
- Direct from your ERP. General ledger, trial balance, master data and budgets arrive through PlaidCloud’s direct ERP connections. An Entity crosswalk maps each source’s company codes onto the same entities, and files fill in anything a system doesn’t hold, such as currency rates or allocation drivers.
- One model, one currency. Load conforms every source into the Common Finance Model and translates it into your reporting currency: profit and loss at the period’s average rate, the balance sheet at the closing rate, and equity at the historical rate. Hierarchies are built from your master data.
- Check readiness. Checks run for each entity and period: the trial balance and sales tie to the general ledger, every product, customer and account is known, currencies translate, and every entity is mapped.
- Many sources, one model. Add, order, edit and remove sources, each an ERP system or instance or a file. Load conforms all of them, keeps one copy where they overlap and records the rows it held back with the reason, and Apply changes re-runs the conform without pulling from the ERPs again.
- Map codes across sources. When sources share a code, a proposal by name match asks whether it means the same thing; answer one code, or a whole group with Accept all, All the same or All different. Mappings are kept for each source and dimension. Up to ten extra categories carry the names you give them, and a coverage view lets you choose which source counts from which month.
- File options. A file can apply to one source, ask the questions it needs answered, split a column into several fields, set the reporting currency, or be removed.
- Management consolidation. Each load builds the group profit and loss, year to date too, with intercompany eliminated on an Eliminations entity, an Unmapped bucket and a reconciliation to your consolidation system’s results. It is a management view that follows the project’s row access, and not a statutory consolidation.
- Keep it current. Refresh nightly reloads the model every night, and a missing or empty file finishes with a warning, not an error.
Start a project from a template, such as Enterprise Profitability, and the project’s Foundation tab takes you from connected ERPs to a loaded, checked model. See Data Foundation: The Common Finance Model.
-
Ask Weave whether you’re ready to close, and check that two tables tie out. No AI is involved in either.
- Close readiness. Name the workflows a close, a forecast or any deadline depends on as a readiness set, from Readiness on a project’s Weave tab. One checklist answers for all of them: “11 of 13 ready; 1 not ready; 1 couldn’t check.” A workflow is ready only when every check Weave could run on it passed, and anything Weave couldn’t judge shows as Couldn’t check with the reason, never as ready. See Check Readiness Before a Close.
- Reconciliations. Compare two tables, or two periods of one table, row by row on the key columns you pair. Weave says whether they tie out, how far apart they are and which rows differ, and never says “ties out” for data it couldn’t read in full. A second Architect approves each reconciliation before it’s used. See Check That Two Tables Tie Out.
-
Try a driver what-if on an Allocation Split step. Right-click the step and choose What if… to see how it would share its cost with different driver values, a different column of its driver table, or a numeric column from another table with the same keys. Compare up to 10 scenarios, each calculated to the cent from the step’s last run, and see who gains, who loses, and which lines would get nothing. Break down by shows which accounts or other values each member’s change falls on, and Later steps carries the scenario into the tables built from the allocation’s output. Nothing is re-run and no data changes. See Try a Driver What-If.
Your Own Warehouse, Lakehouse and Storage
Section titled “Your Own Warehouse, Lakehouse and Storage”-
Run PlaidCloud on your own Snowflake or Databricks warehouse. A workspace administrator opens Use your own warehouse in the user menu and runs a generated setup script in their own account. The script creates a capped warehouse and a service identity for PlaidCloud, and never touches your existing warehouses. Check connection explains any problem and its fix in plain words. New projects then store their tables and run their workflows in your warehouse, and removing the warehouse revokes PlaidCloud’s access there. See Use Your Own Warehouse.
- Sign in with Snowflake or Databricks. Instead of pasting a credential, an administrator can attach the warehouse by signing in. Dashboards then run with each person’s own warehouse permissions, and workflows run as a service identity holding only the data role you name. See Sign In With Snowflake or Databricks.
- Connect more than one. Make default in Manage warehouses sends new projects to another warehouse, while existing projects stay where they are. See Manage Warehouses.
- See what a run costs. Run history shows a Warehouse cost column in credits or DBUs for each run. See Warehouse Cost per Run.
- Your governance stays in place. Descriptions are written to the warehouse’s own catalog, and Snowflake row access and masking policies and Databricks row filters and column masks stay attached when PlaidCloud replaces, recreates or clones a table. See Your Policies and Descriptions Carry Over.
-
Moving a project to Lakehouse v2 is guided from readiness to undo. Check Migration Readiness reports tables, views, steps, dashboards, data shares and capacity as clear, needs review or blocked, with a plain explanation of every finding. Try on v2… makes a private trial copy with schedules off and ERP posting in simulate mode. Start a migration now or at a time you choose. The project’s schedules pause while it runs and are restored afterwards, and you’re told by message and email when it finishes. What Changed… reports each table’s outcome, and Undo Migration… returns the project to Lakehouse v1 within the undo window. See Migrate a Project to Lakehouse v2.
-
Copying a project copies its data, with a choice of how much. Copy Project offers Structure only, Full data or Sample, within one Lakehouse or from one Lakehouse to another, after a review of its size and any warnings. A copy with data is private to the person who made it. Copy Results lists each table’s outcome, Retry failed tables copies only those, and View Import or Copy Result reopens the window later. Sensors and scheduled events arrive turned off on the copy, so nothing fires until you turn them on there. See Copying a Project.
-
See your storage, and control how much history you keep. See Managing Storage and History and Time Travel.
- Workspace storage. A workspace Storage window shows a meter with included, reserved and on-demand bands, the projected monthly average that billing uses, and each project’s current data and history. A Deleted tables awaiting cleanup row shows storage from deleted tables that is still being released.
- Projects and tables. The Projects list has a Size column, and each project’s Tables tab shows Current data, History, Time Travel Days and Note columns and a Storage button that suggests ways to free space. Sizes reflect what is actually stored, history included.
- Time travel. Time travel defaults to 1 day: set a project’s default from 1 to 90 days, and lengthen it for any table. History older than the days a table keeps (its own setting or the project default, whichever is longer) can’t be restored and is cleaned up nightly, so raise the setting for anything you need to keep longer.
Metrics and AI Assistants
Section titled “Metrics and AI Assistants”-
Agree once how a business number is calculated, and every AI answer uses that calculation. A project Architect defines a metric, such as Net Revenue or Headcount: its source table, how its rows add up, the rows that always count, the lookup tables it joins, and the columns it can be broken down by. A definition is live as soon as it’s saved, and a save that names a missing table or column, or a lookup with more than one row per key, is refused. Ask an AI assistant for the metric by name or synonym and the figures are calculated by the definition, under your own row access, not worked out by the assistant. Definitions are managed through an AI assistant or the REST API. See Metric Definitions.
- Derived metrics. Define a ratio, sum or difference of two metrics, such as net revenue per FTE, or a metric times a fixed factor. See Derive a Metric From Others.
- Compare, rank and share. Ask for the variance against budget, the top or bottom N with ties kept, or each row’s share of the total. See Compare, Rank and Share.
- Your words, the agreed values. Bind a breakdown column to a dimension, and a word such as EMEA becomes every region beneath it. A word that matches nothing, or more than one member, comes back as candidates to ask you about, never a guess. See Ask About a Metric.
- Fiscal time. Tell a metric where its rows sit in your fiscal calendar, and ask for a fiscal period, year to date, year over year or period over period. See Compare Over Time.
PlaidCloud’s own AI assistant answers a business number from a metric first, and shows the figures in their own card marked with a ✓, which never attaches to the assistant’s own text. See Business Numbers and Figures From a Metric.
-
Build charts and dashboards from your AI assistant, with no separate connection. Connect only PlaidCloud’s MCP connector and ask for a table, bar, line, pie or big-number chart over a published table or view, then put charts on a dashboard. You get a link and a sample of the data to check, and you can ask to edit a chart you own. See Build Charts and Dashboards With an AI Assistant.
-
AI assistants and Microsoft 365 Copilot know more about your data and do more with it.
- Guidance for each analysis path and column. Each analysis path has Prompt Guidance, up to 1,000 characters, for the filters that always apply, the fiscal calendar and your business terms. A column memo reaches assistants and LLM steps with the table’s columns, so a code or flag is read the way you meant it. See Making a Path Useful to AI Assistants.
- Copilot does more. The PlaidCloud agent in Microsoft 365 Copilot creates CSV, Excel, PowerPoint and Word files, answers business figures from your metrics, totals breakdowns over the whole table, and explains how an allocation works. Administrators download the updated package and update it in the Teams admin center. See Microsoft 365 Copilot.
- More done in one call. An assistant can group and total a table on the server, with row access applied first, search and read PlaidCloud’s documentation, create a fully configured Document account, including OneDrive and SharePoint, with write-only credentials, remove labels as well as add them, and work with the views you see in PlaidCloud. See Analysis Paths and Getting Started With AI Agents.
-
Weave watches more of what can go wrong, and tells you between runs. Weave’s checks use no AI. Every finding, and every answer from the AI assistant, has a How Weave knows panel: for a finding, the check and the values behind it; for an answer, each query the assistant ran, with every sentence marked Matched, Unmatched or Not checked. See What Weave Checks Today, How Weave Knows and How Weave Knows in the AI Assistant.
- Tables written on time. Weave reports a table that hasn’t been written since its schedule came due, or for longer than a freshness threshold you set under Freshness in the table’s details, from 1 hour to 180 days.
- Inputs that are stale or broken. Weave reports a workflow that reads a table another workflow has written since, and every workflow and dashboard that rests on a failed run, including a dashboard whose charts are built on SQL datasets. Sync Dependencies adds the connectors missing between a step and the table it reads. See Sync Dependencies.
- Data that changed shape. Weave reports a decimal column whose values spread very differently after a run, a join that starts writing rows more than once, values an Extract step left empty because they couldn’t be converted to their column’s type, and an import that found no files. It also reports a metric whose latest period is far from usual, naming the item that drove it. Weave allows for month-end and weekday patterns, so a load that’s always heavy at month end isn’t reported. Go To Metric on a metric finding shows the metric’s figures for the item and period Weave judged, calculated when you click it under your own row access, and Go To Rows opens the rows behind the change. See Unusual Metric Totals and Go To Metric.
- Runs that got slower. Weave names the step that made a run slow, and reports a workflow whose newer runs take at least a quarter longer than its older ones, naming the steps whose time grew most.
- Told between runs. An hourly check tells each problem’s owner once, in an in-app message, and a Weave button in the header counts the problems you haven’t read across all your projects. See Being Told Between Runs and The Weave Button.
- Owners, comments and expected warnings. Each finding has an owner, by default whoever last ran its workflow. Assign hands it on, comments record progress, and the Team view shows who owns what. An Architect can mark a warning This Is Expected, with a reason and an end date. See Give a Finding an Owner and Mark a Finding as Expected.
-
The fix assistant explains an error, proposes a change you approve on a card, re-runs the step and can undo the fix. For the most common failures, such as a column that isn’t in the table, an empty required setting, SQL the database can’t read or a value of the wrong type, the first reply explains the usual cause and how to fix it. A proposed change appears as a card that states why the step failed and shows each changed setting with its current value in red and its new value in green. Approve applies it, and the assistant then offers to re-run the step and diagnoses a new failure. Fix on a finding in the Weave tab opens the assistant on that problem. Undo restores the step’s earlier settings, with Undo & Run to rebuild its output. See Fixing Workflow Errors with AI, Fix a Finding with the AI Assistant and Undo a Fix.
-
See everything Weave did in a project, and export it for your auditors. Activity on a project’s Weave tab lists the problems Weave found and later stopped finding, every change the AI assistant proposed, who requested and who approved or rejected it, exactly what it replaced and wrote, and how it ended. Weave records your approval before it makes a change, and makes no change it can’t record. Export CSV or Export PDF saves a period as audit evidence, and entries can’t be edited or deleted. Summary shows a month at a glance: problems found and resolved, approved changes and how many were undone. Workspace administrators get the same for every project from Workspace Summary…. See Weave Activity History and Weave Monthly Summary.
-
Know what a step does, and what changing it affects, before you save. All of these are built from your project’s saved settings, with no AI involved, and anything your role hides is counted, never named.
- Documentation for every step and workflow, and columns for every table. A Documentation section in the details panel describes what a step does and a workflow’s steps in order, and what reads and writes a table is available to Weave and AI assistants. A table’s details panel lists its Columns with their types and descriptions. Download Handover Pack saves the whole project as one document, metric definitions and allocation models included. See Documentation for Tables, Steps and Workflows.
- What a change affects. Above Save, Weave says how many steps downstream read what the step writes, such as “Changes to this step affect 8 downstream steps in 2 workflows,” and Show the steps lists them. See What a Change Affects.
- What Save would refuse. A strip above Save runs the step’s checks as you edit, with Go To for each problem, and the Expression Editor checks an expression as you type. See What Save Would Refuse.
- A review of one workflow. Check on a workflow’s canvas or Steps grid runs Weave’s checks on it, plus allocation and row-level security checks. See Reviewing One Workflow.
-
Turn off any of Weave’s AI features for your workspace. Workspace administrators choose AI Features… on the header’s Weave button to turn Weave chat, the fix assistant, the import and export assistants, the expression assistant, SQL from a description, the workflow review summary or the starter chart on or off, each on its own. Everything is on until an administrator turns it off, Weave’s checks and findings always stay on, and each change is recorded in the Security Log. See AI Features.
Panel Apps
Section titled “Panel Apps”-
Build, publish and brand Panel apps inside PlaidCloud.
- Panel Studio. New Studio App, the first button in My Panel Apps, starts from a blank app, a table, a view, another Studio app or the app gallery, and Create Panel App on the Tables tab starts from the table you select. Open in Studio opens the app’s files in an editor. Save a draft, Preview a private build, and Publish the live app. Build errors point to their file and line, History restores any published version, and Insights shows views, load times and errors. Ask Weave knows the app’s files and build status, and shows you the full difference for approval before it saves any code. See Panel Studio and Build Panel Apps With Weave.
- An app kit that looks like PlaidCloud. Apps built with the kit use PlaidCloud’s colors, type and spacing in light and dark. The kit includes navigation, filters, headline-number cards, charts, forms, a command palette, buttons that run a workflow, printing, and a data grid that sorts and totals on the server, formats numbers per column, and loads one screen of rows at a time, so it handles any table size. Apps read tables as the signed-in viewer, with row security applied. See The Panel App Kit, DataGrid and The Data API for Panel Apps.
- Alerts and digests. Get told by push notification, email, Slack or Microsoft Teams when a number from a project table crosses a limit, or get a daily or weekly digest of up to 10 numbers, always under your own row access. See Alerts and Digests.
- Projects, branding and sharing. Assign an app to a project, after a review of who can open it but isn’t in the project. Edit Branding on a project’s Home tab sets eight brand colors and a logo for every app in the project. Gallery… lists a Studio app that belongs to a project as a template for everyone who can view its project, and Take out of gallery removes it. See Project Branding and App Projects and The App Gallery.
Workflows
Section titled “Workflows”-
Optimize an Advanced workflow to run faster. Right-click a workflow and choose Optimize… to review changes before any happen. Rebuild dependencies from data lineage connects each step only to the steps whose tables it uses, so independent steps run in parallel, and Merge combines a chain of Extract Data steps into one, applied only after both versions give identical results on your current data. The window shows the time saved and lists each step it couldn’t optimize with the reason, and Undo optimization… restores the workflow. See Optimize a Workflow.
-
Arrange and edit the Advanced workflow canvas more directly. Hold Cmd or Ctrl and drag steps into or out of a container, and wire an execution container’s output from its Exit port. The Arrange menu tidies, aligns, distributes and snaps steps while keeping your order and moving notes with their steps, each as one Undo. Right-click empty space to copy and paste steps where you clicked. Create Step on a selected step starts a join or extract from the table it writes, and a window with unsaved changes asks before discarding them. The eye icon on a step’s input or output opens a preview of exactly that table. See Containers, Arrange the Canvas, Copy & Paste on the Advanced Canvas and Create a Step.
-
Get a join’s keys right before you run it. Guess on the Join Map uses the keys of another step that joins the same two tables, then a metric, then the tables’ key columns, then matching names, and says which it used. Check Keys counts key values that repeat in the table being joined to, which can multiply your rows. When Inspect Source fills a step’s columns, an identifier column such as
customer_idstarts as Group By rather than Sum. See Inner Join. -
The REST Request form helps you build and check a request. Filter a large endpoint catalog, or paste or drop a Postman collection, OpenAPI spec or HAR file. Headers and parameters are checked for duplicates, mistyped names and secrets typed in plain text, and a test request shows a formatted response and a suggested items path. The step adds page-number, cursor and Link-header paging, proxies, Digest and NTLM sign-in, SOAP calls, fractional request rates, polling for a running job, and a raw response mode that keeps each error as a row with credentials masked. A POST, PUT or PATCH is retried after a rate limit but never after a server error. See REST Request Step and REST Request.
-
New workflow steps.
- Files and documents. Import Delta Lake, Import ORC and Export to ORC, Export to PDF, which writes a table to a PDF, Update Excel Cells and Document Write Blob. File imports and exports read and write Zstandard, Snappy, LZ4, LZO and Deflate files.
- Encryption. Document: Encrypt File and Document: Decrypt File use AES or Blowfish, or set or remove an Excel workbook’s password, with the key always held in a variable.
- Tables and sources. Table: Restore to Timestamp puts a table back to its newest snapshot at or before a time you give. NetSuite: Import SuiteQL Query loads a SuiteQL query you write into a typed table, and KNIME Rule Engine applies rules read from a rules table.
- Spatial. Spatial Reproject and Spatial File Export.
- Machine learning and text. ML: Train Model adds XGBoost, gradient boosting, support vector machines, clustering and PCA, ML: Score adds class probabilities and model reports, and Table: Text Function adds string distances and JSON and XML functions.
-
More options in the steps you already use. Every option is a field in the step’s form.
- Imports. Import Excel reads a different set of worksheets from each workbook and opens password-protected workbooks. Import Fixed Width reads EBCDIC and Latin-1 files, and Import XML reads one repeating element into a table. File searches accept wildcard and regular expression patterns.
- Exports. Export CSV and Excel can append to an existing file, most exports can stop instead of replacing a file that’s already there, and CSV, Excel, JSON and SQL exports can sort their rows. Export to Excel can write a sheet or a file for each value of a column, and Export to CSV can escape or omit quotes the way Spark does.
- Tables and flow. Table Extract and the joins can carry columns named at run time, so a step after a pivot needs no column list. Table: Union All can stack sources by position. A step can run depending on whether a table exists (Table Exists Conditions), Workflow Loop can carry on past a failed iteration, and Wait can wait for a time of day or a file change.
- Notifications, AI and remote SQL. Notify Via Email adds Cc, Bcc and Reply-To, and can send each row of a driver table its own email, and Notify via Slack can attach a file. Text: NLP (AI) can write embeddings and reports each failed call. Remote Execute: SQL can run a statement, such as a stored procedure, once for each row.
-
Bring Savant, Prophecy and KNIME work into PlaidCloud. Choose Tools → Import Savant Analysis, Import Prophecy Project or Import KNIME Workflow in Analyze, or use the REST API or an AI assistant, to turn an export into runnable workflows built from standard PlaidCloud steps, with the graph preserved (Savant notes included) and parameters turned into project variables. Bind each source to a project table or a Document file, and each connection the project reaches to one of yours. Credentials are never carried in a workflow, and a conversion report names anything left to set. See Migrate Savant Analyses, Migrate Prophecy Projects and Migrate KNIME Workflows.
-
Workflow texts go to members who opted in, and are confirmed. Notify: SMS texts workspace members or a distribution list. Each member opts in from Text alerts in the user menu with a confirmed US mobile number. A text that can’t be delivered is reported on the step and sent by email instead, repeats are collapsed, and quiet hours hold texts until they end. A step still set up for the retired carrier-gateway texting completes with a warning and sends nothing until you choose recipients. See Notify Via SMS and Text Alerts.
Connections and Tables
Section titled “Connections and Tables”-
Connect to MongoDB, Apache Kafka, Google BigQuery, Azure Cosmos DB, Smartsheet and Tableau.
- MongoDB and Apache Kafka. MongoDB and Apache Kafka connections come with Import MongoDB, Export MongoDB, Import Kafka and Export Kafka steps. Kafka signs in with a user and password or with OAuth for services such as Confluent Cloud.
- Google BigQuery. A Google BigQuery connection signs in with a service account key, for both Import SQL and Export External Database Tables.
- Azure Cosmos DB, Smartsheet and Salesforce. A REST connection authenticates to Azure Cosmos DB with its master key or Microsoft Entra ID, a Smartsheet Document account attaches files to a sheet or imports a file as a new sheet, and Salesforce: Query Object can include deleted and archived records.
- Tableau. Publish to Tableau creates, overwrites, appends to or upserts a data source on Tableau Server or Tableau Cloud.
-
Find any table by its column, its value or the workflow that uses it. The Tables tab has a By Workflow folder, a column search with a Column Tree of the tables that feed and are fed by each match, and Search values in data… (slow), which lists each table and column holding a value you type, within the rows you could already open, and opens a result in the Table Explorer. Table Details shows a table’s Published As name with a copy button, and on a Snowflake or Databricks warehouse the delete window shows when a table was last read and by whom. For Document accounts, Where Used lists every step, workflow and project default that uses an account, and Re-point… moves them to another account after a preview. See Finding Tables, Search Values, Last Read and Document Account Where Used and Re-Point.
-
Turn a CSV into a table in any project, and see it at a glance. Drop a CSV onto a project’s Tables list, or choose New > Table from CSV File, and PlaidCloud imports it as a new table. New > Project from CSV File creates a project from one. Your data at a glance then shows the row count, each column’s type, its share of empty values and distinct values, and a starter chart, with a starter dashboard one click away. See Import a CSV File and Start With Your Own CSV.
Dashboards and Administration
Section titled “Dashboards and Administration”-
Choose which security groups can see each dashboard, in one dialog. Who can see this? sets a dashboard to Everyone in the project, Specific groups or Only owners (Draft), with a preview of who gains or loses access. Someone blocked from a dashboard gets a Request access page that emails its owners. Workspace admins get a Dashboard Access Review, and a project’s Default Dashboard Audience applies to its new dashboards. An audience hides the dashboard, while chart data still follows project access and Row Access. See Control Who Can See a Dashboard.
-
Set up single sign-on with your own identity provider, on paid plans. Single sign-on on the Workspaces page adds Microsoft Entra ID, Okta or any OpenID Connect provider, with a copyable redirect URI and a Test sign-in button. People with an existing account link it by signing in to that account once. Password with MFA and social sign-in with Google, Microsoft and GitHub remain on every plan. See Single Sign-On.
-
See who uses what with Usage Reports. From the Identity tab, administrators with the Security Log permission see active, new and returning users, the adoption of each screen, dashboard, Panel app and notebook, the most common paths between screens, and a funnel of up to six steps. Choose a range of up to 180 days, download summaries as CSV, and erase a member’s usage on request. See Usage Reports.
Work Anywhere
Section titled “Work Anywhere”-
Run and monitor your workflows from your phone, and ask Weave on the go. The free PlaidCloud app is on the App Store for iPhone and iPad and on Google Play for Android. Sign in on the sign-in page of your workspace, single sign-on included, and browse your projects, follow each step’s state and run history, and Run, Pause, Resume or Stop a workflow. Ask Weave about your workspace, or tap Ask Weave on a failed step to start with its context. A change Weave proposes waits for you to tap Approve and confirm with Face ID or your fingerprint. See The PlaidCloud Mobile App and Weave on Mobile.
-
Work with PlaidCloud data in Excel with PlaidXL. The add-in signs you in with your usual sign-in, and each workbook remembers its own project and workspace. See Work in Excel With PlaidXL and the PlaidXL reference.
- Retrieve and refresh. Retrieve a table’s chosen columns and rows into a worksheet, and Refresh All Tables brings them up to date.
- Save back to PlaidCloud. Edit a retrieved table and Save to PlaidCloud… writes it back after a confirmation, with a restore point and a Restore previous version link. Empty cells and cells showing an error save as empty, and a filtered table or columns you didn’t retrieve each get an explicit warning first.
- Hierarchy formulas.
PLAIDXLformulas lay out and drill a hierarchy, and calculate a measure rolled up to any combination of dimension nodes. A report full ofCELLformulas needs fewer lookups per cell, a cell that can’t be calculated shows the reason on hover, and the Formulas tab builds a formula for you. - Share a workbook. Convert to values (for sharing) and Restore PlaidXL formulas on the ribbon prepare a copy for people without PlaidXL, and the task pane follows your Office theme and works from the keyboard.
-
Use PlaidCloud in your own language. Language in the header offers English (US), English (UK), Deutsch, Español, Français, Português (Portugal), Português (Brasil), Polski, 日本語, 한국어 and 简体中文. Your choice follows you to other devices. Numbers and dates follow the language, while values you save or export are unaffected, and Japanese, Korean and Chinese input methods work in number fields. Emails, sign-in screens and error messages follow your language too, and this documentation is available in all of them except English (UK): choose one from the language menu at the top of any page. See Change Your Language.
Changed
Section titled “Changed”Changes That May Need Action
Section titled “Changes That May Need Action”-
Every Lakehouse stores and calculates the same way. An empty field from Import CSV, Import XML or Import Table Archive loads as a null, so a later step that tests for
= ''or= 0needs to test for null instead.cast(x, Float)is 64-bit everywhere, so large amounts keep their digits, and Float columns PlaidCloud creates are double precision. JSON columns, Table: Delete Rows and conditions that aggregate work on every Lakehouse. Steps that build their own results, such as Solver, Forecast, and Pivot and Allocation Split when they run in pandas, load them as typed data, much faster. Table Explorer shows the minimum, maximum, mean, sum, standard deviation and variance of a numeric column. See Casting to Float and Column Data and Unique Counts. -
User-defined functions run on Python 3.14 with newer libraries. Most functions run unchanged. A function that imports numba, or a standard-library module Python has removed, such as
cgiortelnetlib, now fails and needs rewriting. Forecast’s AutoTheta bounds can differ slightly from earlier runs. See User Defined Transform. -
Full agent access is deprecated and will be removed on January 15, 2027. An LLM step using Full access logs a warning. Move to Read & write, where agents change data through PlaidCloud’s write tools and an LLM step writes only to bindings with Write checked. See Agent Access.
-
Document accounts are easier to set up and browse. A OneDrive or SharePoint account’s file browser opens the folders under its start path. A Google Cloud Storage account needs a valid service-account key, and an existing one without it stops working until you add one. A rejected setting explains itself, and the unused Backup icon is gone from Manage Accounts. See Add Google Cloud Storage Account and Backing Up Document Accounts.
-
The API call that deleted every table no step referred to is removed. Delete tables by ID instead.
Improvements
Section titled “Improvements”-
One Weave tab holds your findings and your Weave AI chats. Chats are tabs inside the Weave tab, after Findings. Each finding is a card with its own Fix, Go To and Mark Read buttons, a dashed card marks something Weave couldn’t check, and chips filter the list by what needs attention. A finding about an expression names where the error is, and the expression itself is under How Weave knows. Weave mentions each problem once when runs end, and messages about one project join into one rather than stacking up. The tab opens quickly in large projects because Weave checks again only what has changed. The Weave column is gone from the workflow list. See The Weave Tab and Being Told When a Run Ends.
-
Ctrl/Cmd-K opens the command palette from any screen. Before you type, it lists what you opened last and the actions you can take: go to Home, Projects, Dashboards or Documents, switch between light and dark, connect your AI, or get help. Typing narrows both lists, and from two characters on it also searches projects, workflows, tables, dashboards and Panel apps. See Search and Commands.
-
Steps catch problems when you save, not when they run. Saving a step whose settings would fail its run, or give the wrong result, is refused with a message saying what to change. Each step’s reference page lists what it checks. A mapped source column that the step or its source table no longer has is flagged in red, and saving opens a dialog with Remove All and Save, Save Anyway or Cancel. Join, lookup and spatial steps keep such rows in place instead of deleting them on their own. An expression that uses an undefined variable names the variable, and a step refused when it runs explains why in plain words. See Source Columns That Are No Longer Listed, Columns Missing From the Source Table and Managing Step Errors.
-
A missing or empty source is never an error. A file import whose file is missing or empty finishes with a warning naming the path and leaves its target table empty with its declared columns, so later steps read zero rows. Document steps over a missing or empty folder report “Nothing to process” and the workflow continues. Remote file steps warn on a missing path and refuse an unsafe one. Checks for a missing file work on OneDrive, SharePoint and SFTP. Report: Generate Batch with an empty batch table finishes with a warning and produces no reports. See Import Steps, Document Workflow Steps and Document Path Conditions.
-
Large workflows run reliably, and Stop cancels imports and exports. When PlaidCloud is busy, steps wait for capacity instead of failing, and capacity is added as steps queue up. Import External Database Tables runs up to 10 tables at a time. If the server running a step is lost, a step that replaces its output runs again automatically, and one that can’t be repeated safely fails once with an interrupted message. Stop cancels running import and export jobs, and a stopped run reports Stopped. See Interrupted Steps and Pause, Stop, and Resume.
-
File exports keep their types. Parquet files keep whole-number columns as integers, store timestamps in microseconds and label a time zone as UTC. Avro files keep dates, timestamps and booleans typed, and a single-table Avro export is named after the export. Choosing many files for a Parquet export stops with a message that Parquet writes a single file. Very large Parquet, ORC and JSON exports are written as the rows are read, so they no longer run out of memory, and Parquet uses the compression you choose. A PlaidLink agent on version 1.2.33 or later receives an exported table as a Parquet file. See Export to Parquet, Export to Avro and How the Data Reaches the Agent.
-
ERP posting follows each schedule’s mode, and an uncertain answer is held for you to check. A scheduled event keeps its own posting mode, Live, Simulate or Off, capped by the project’s setting, and Run Selected Events uses each event’s mode. An ERP answer PlaidCloud can’t confirm, such as a timeout or a server error, is held as in doubt rather than recorded as rejected, and a posting can’t be reversed twice. The Posting Register lists reversal outstanding postings, and an Architect authorizes a project to post from ERP Write Authorizations on its home page. See ERP Posting Mode, Reversal and Authorize the Project to Post.
-
Row access keeps up with group changes, and says what’s wrong. A group membership that single sign-on sets when someone signs in reaches dashboard row access within about 15 minutes, and a new member of your default groups gets their rows within about a minute. The Row Access window warns when a restriction filters nothing, handles column names containing
&,/or%, and shows Error with the reason or Retrying with a Retry button when a sync fails. See Keeping Dashboards in Sync. -
The Project app handles invoices, trips and late time more completely. A new invoice starts from a master project, which supplies the customer and PO number. Deleting a draft invoice makes its time and trips billable again, and AR Recipients sets who is emailed when an invoice is submitted. The trip drawer shows each line’s vendor, category and audit status, and Add Late Time starts a sheet for an approved week. See Track Time and Expenses with the Project App.
-
Smaller changes around the app. Double-click a row in an account, dashboard, member, group or setup list to open it. Empty Dimensions, Data Editors, UDFs and Steps tabs explain what to do next, and a member with no apps lands on the Launcher. A step’s table selector shows when the table was last updated (Seeing When a Source Table Last Changed). Export External Database Tables maps columns on its own tab, and Report: Generate Batch no longer shows the unused Report Constants section. Convert Encoding recognizes Central European, Turkish and Arabic encodings. A link to a removed screen says so.
-
The AI assistant changes nothing until you approve, and shows names as written. Building a step into a new table created the table before you approved; now the table and step are created together after you approve, and creating a project, workflow or dimension, copying a table, and changing table properties or labels ask first too. An approval counts only in your next reply. An import or export the assistant proposes shows its table, document account and workflow by name, and is never offered with one that doesn’t exist or that you can’t open. Names with underscores display exactly instead of turning into italics, tables and code are readable in the Dark theme, diagrams draw instead of showing as text, and the Expression Editor’s Weave panel reads the expression you’re editing. The assistant lists a project’s tables even when one has no name, and answers which steps read or write a table even when a step reads or writes several tables, instead of stopping with an error. See Building a Workflow Step.
-
File imports read what you point them at. Import Excel reads whole-column ranges and XLSB workbooks, finds sheets in a zipped workbook, and no longer fails on headers with padding or line breaks. Import CSV honors Skip Rows, so a file with lines above its header, such as an SAP list export, imports by column name, and it runs with default delimiter, quote and escape characters. One source column can feed several target columns, an import whose columns are all fixed values loads every row, a file named with an uppercase compression extension is decompressed, and a skip-when-unchanged import reloads a table that was emptied. Asking the AI assistant to import a file from Project Default reads the project’s default document account, and tells you plainly when the project has none set. Parquet files with newer column types open in Documents. See Import Excel and Import CSV.
-
Workflow steps save and run as configured. Table: Upsert steps save again, with their Update Key column back. Table: Project to Project Copy names its target project and reads through data shares. Document: Save Dashboard as PDF keeps its column types, and Table: Union Distinct and Union All match columns by name. Multi-Table Join saves with several inputs and accepts column names with spaces. Table Lookup ignores aggregation settings on its lookup columns, and SCD-2 As-Of Join takes a variable as its date. LLM steps run on Claude models older than 4.6 and keep their model when an AI assistant edits them. REST Request sends row values encoded once, and exports to a remote database commit their rows. Steps that pick a workflow, such as Workflow Delete, Copy, Rename and Stop, keep one that no longer exists, shown as id (missing), instead of switching to the first workflow in the list. See Update Key, Data Sharing Management and Multi-Table Join.
-
Workflow runs finish as they should. A Workflow Loop runs the looped workflow on every iteration, as do iterating containers and workflows run more than once in a run. A run that starts at a step, or inside a parallel group, runs the right steps and finishes. A long-running import or export is never started twice, doesn’t stall during a PlaidCloud update, and an action that ended in a server error isn’t sent again. A database driver crash fails only its own step, with the database’s reason. Run history counts the rows of every step, and a step that fails before its work begins writes the reason to the workflow log. See Workflow Loop and Run History.
-
Projects copy, import and open without surprises. A project import keeps everything it could restore and lists what it left out in an Import Result window, shows its progress, says why a failure happened, and can be retried. Copying chosen objects works when the project has variables. Opening a project no longer changes its settings, Bring your own data imports your CSV, and creating a table on a PostgreSQL warehouse works. App Runner apps refresh themselves when PlaidCloud updates and work for viewers on a project that restricts rows. See When Part of an Archive Does Not Import and Keeping the App Current.
-
Dimensions and allocations load and read correctly. A parent-child load keeps rows with an empty parent as top-level members, and a member outside the main hierarchy is refused for a second alternate hierarchy in every kind of load. Moving a member under a parent its hierarchy doesn’t contain is refused, a dimension step names the dimension it couldn’t find, and pasting a column with blanks keeps each value on its row. Reading a large dimension no longer holds up everyone else in the workspace. An allocation rule whose dimension property leaves Hierarchy as Not Specified runs. See Load to Alternate Hierarchy and Dimension-Type Properties and the Member Picker.
-
The workflow canvas and screens behave. Drill to Workflow opens an Advanced workflow on its canvas, and the canvas opens every workflow, telling you when part of a layout couldn’t be read. Toolbar buttons that don’t fit move into a More menu, and lists update as soon as you save. Notifications show their whole message, warnings show the warning icon in the Log tab, Table Explorer says when a column’s profile failed, and a Data Editor keeps empty date cells empty. Saving a Data Editor with a column filter on keeps the rows the filter hides, including after you add, delete or move rows. The user-defined function editor opens editable for anyone allowed to change it. Deleting tables from the Tables list lists any step that reaches a table only through a variable path, such as
/Periods/{period}/Sales, and deletes that table only if you force it. See When the Canvas Can’t Open a Workflow and Delete. -
Members and connections work as they should. Add Workspace Member works for someone who already has an account, including one created by single sign-on, and never lowers a project role they hold. ODBC connections connect through a DSN or driver on the PlaidLink agent’s machine. A provider that rejects your OAuth credentials tells you to re-enter or re-authorize them, Databricks samples and exports refresh an expiring token, and connections of older database types can be deleted. Where Used includes notification steps, and free workspaces keep daily, weekly and monthly backups. See Member Management and ODBC.
-
The REST API and AI assistant tools do exactly what you ask. A partial update to a member, group or scheduled event changes only the fields you name, so a paused event stays paused. Starting a run at a named step starts there, and restoring a table snapshot keeps a safety snapshot and an audit record. A driver what-if from an AI tool answers only where it agrees with the split your model actually made. Requests that used to fail with a server error now work or explain what was wrong. See Tracing Allocations and Reverting or Restoring a Snapshot.
Security
Section titled “Security”-
AI assistants stay within your permissions. The assistant’s data queries are checked by their structure before they run, and only a single read query runs. A read-only AI connection, including Microsoft 365 Copilot, can’t export files, save dashboard PDFs, copy rows into another project, or send changes through a REST connection. An import through an AI assistant runs as you, in the project you name, and a conversation can be continued only by the person who started it. The assistant and table references leave out tables, steps, workflows and dimensions your role hides, and Weave counts a failed step you can’t open rather than naming it. See Access Levels, What Your Role Hides and Manage Past Conversations.
-
ERP posting can’t bypass its controls. With segregation of duties on, reversals go through the Posting Register, where approval applies. A copied or imported project no longer inherits permission to post, SAP and EBS postings can’t be sent to an agent directly, and an SAP Post step in Test Only Mode needs the same permission as a live posting. An undocumented endpoint that posted to Sage Intacct without these checks is removed. Error messages from ERP and REST connections leave out credentials. See Segregation of Duties.
-
Secrets are write-only, and only owners decide who uses a connection. Document account keys, passwords and tokens are never shown again once saved, and changing where an account connects needs the secret again. Passwords and keys in steps live only in variables; secrets already typed into steps were moved into project variables visible to Architects only. Multi-line credentials are masked as you type. Only a connection’s owners or a workspace admin can change who has access to it. See Saved Secrets Are Write-Only, Keys Live in Variables and Create and Manage a Connection.
-
A document account you can’t use stays closed to you. Checking whether a file exists, reading the columns of a Parquet, Avro, ORC or Delta file, filling an import step’s columns with Populate, and the AI assistant’s file inspection now refuse an account you don’t have access to, as browsing it already did.
-
Only a workspace admin can change a workspace admin. Editing, deactivating or deleting an admin, changing their groups, and making someone an admin all need a workspace admin. See Changing a Workspace Admin.
-
A data share speaks only for members of the project it was granted to. Reading through a share needs a role in the receiving project, and writing needs a Write share and the Manager or Architect role there. See Sharing Data Between Projects.
-
Restores apply only to the item named. Restoring an earlier version restores the item itself, and the permission check is on that item. A dimension backup file whose contents don’t belong to the dimension in its own header is refused. See Version History.
-
Templates, imports and archives refuse unsafe content. Export templates and report templates render in a sandbox that reads only the data they’re given. An import builds its loading query from its own settings, a project archive holding content a PlaidCloud archive never contains is refused, and a project slug that would share another project’s reporting schema is refused. See What a Template Can Use and Restoring an Archive.
-
Dashboards follow PlaidCloud sign-in and your row access. A dashboard’s thumbnail is rendered as you, accounts come from PlaidCloud sign-in only, and dashboards import only from the ZIP file that Export produces. See Importing Existing Dashboard.
-
Starting a workflow run no longer returns the run’s credential.
-
The environment that runs user-defined functions no longer includes PySpark’s Java libraries, which carried known vulnerabilities. PySpark’s DataFrame and SQL API still reaches your Spark Compute Cluster, but a function that imports
SparkContextorRDDnow fails. See Getting Started with PySpark. -
Attachments, restores and conversions read only accounts you can use. Emailing a file, restoring a dimension or a table archive from a file, inspecting or importing a table bundle, filling an Excel export from a template, and converting an Alteryx workflow now refuse a document account you don’t have access to. Backing up a dimension to a document account needs write access to it. Workflow steps are unaffected, except that a step emailing a file, restoring tables or filling an Excel template can no longer read from a Personal Temp Drive.
-
Exports save only to accounts you can write to. Exporting a table or query to a file in any format, exporting a table archive, a project archive or a table bundle, creating local Python workflows from a project, saving REST responses as files, and converting an Alteryx or Prophecy workflow now need write access to the document account they save into. Exporting a query also now checks that you can use its project. A table bundle import no longer writes its import log into an account you can only read. Export steps in workflows are unaffected, except that they can no longer save to a Personal Temp Drive.