July 2026

Releases Shipped

Version	Released
v5.11.0	2026-07-09

Added

PlaidCloud Git connections. Connect to your workspace’s own managed Git server with a simplified form — just an account name, memo, repository path, default branch, and optional start path — with no server URL, username, token, or SSH key to enter, because your Git server and your access to it come from your normal PlaidCloud sign-in. Point a Server Panel app’s build at a PlaidCloud Git connection to automate builds entirely inside your workspace, without an external GitHub account or access token. External Git providers (GitHub, GitLab, Bitbucket, Forgejo, Gitea) keep their full connection form. See PlaidCloud Git Connection.
Skip re-importing unchanged source files — file-import steps (CSV, Excel, Fixed Width, JSON, Avro, Parquet, Alteryx yxdb, dbf, XML, X12 EDI, Access) have a new Skip re-import if source files unchanged option. Turn it on and a workflow run skips the whole download, conversion, and table rebuild when the source files and the step’s settings have not changed since the last successful run — so a scheduled reload no longer repeats the work. On by default — untick the option on a step to always re-import; the change is detected from each file’s size and modified time, plus a content hash where the source provides one. Manual, one-off imports always run. See Import CSV.
Machine learning in workflows — ML: Train Model and ML: Score steps. Train a model on any table — choose from seven algorithms including linear and logistic regression, decision trees, random forests, and gradient boosting — and score data with it, appending a prediction column, entirely inside a workflow. The trained model flows through the workflow like any other table, with its algorithm, settings, features, and training metrics recorded alongside, and mistyped algorithms or settings are caught when you save the step. Alteryx Assisted Modeling pipelines also convert to these native ML steps on import, with XGBoost models carrying a documented approximation note. See ML: Train Model, ML: Score, and Machine Learning Conversions.
PlaidCloud Managed Bucket document accounts. Add fully managed file storage to Document in a single step — no cloud project, credentials, region, or storage tier to set up or tune. Available on the Business and Enterprise plans. See Add a PlaidCloud Managed Bucket Account.
A more capable REST Request step. REST Request can now fan a call out over the rows of a driver table — one request per row, with an optional row filter — carrying key columns onto the results, with a concurrency limit, a request-rate cap, an opt-in idempotency key for safe re-runs, a continue-on-error option, and clear per-row failure reporting. New authoring aids make requests easier to build and verify: import from a curl command, insert and highlight variable tokens (including a driver table’s columns), test the first fan-out row, preview the resolved request and its pagination before running, restore unsaved drafts, edit the selected connection in place, and confirm before saving a request you haven’t successfully tested. See REST Request Step.
An expanded Help & Support experience. Attach screenshots and files to tickets and replies (images show inline), work tickets from a sorted, filterable support console (reply, add internal notes, claim, set priority, and escalate to PlaidCloud), route new-ticket alerts to a Slack, Teams, or email channel, email the ticket owner when a ticket gets a reply, and auto-escalate tickets left unanswered too long. See Getting Help, The Support Console, and Managing Support.
Analysis paths for AI questions. Give the tables you analyze most a friendly name, and set one as your default, so you and your AI assistant can ask about “the Operations Results” — or just “what changed last quarter?” — without naming the project and table every time. Works across Microsoft 365 Copilot and any MCP-connected agent. See Analysis Paths.
Smarter, more honest AI allocation cost-tracing. When you ask an MCP-connected AI agent why an allocation result changed, the answer now carries a plain-language confidence level and caveats — flagging when a cause is a mix of overlapping factors, when it depends on current hierarchy or driver weights, or when part of the change may just be an unfinished data load, and these trust signals travel with the answer even when a follow-up agent summarizes it. Administrators can also turn on an optional check that lowers confidence and flags an answer when the pool or driver data behind it hasn’t been reloaded recently. For a result built from several allocation branches, the confidence now weighs each branch by how much of the change it accounts for, so one small, shaky branch doesn’t drag down an otherwise-clean answer. You can also ask driver-reweight what-if questions (“what if this cost centre’s headcount doubled?”), and the full root-cause analysis can now be driven over the REST API, not only through an AI agent. See Tracing Allocations with an MCP-Connected AI Agent.
Preview step data in the Visual Workflow Designer. See a step’s output in place, without opening a new window — dock a preview of the first 100 rows, with typed columns, row and column counts, and a selector for multi-output steps, at the bottom of the canvas. Clicking another step retargets the preview, it refreshes when a run finishes, and you can open the full Table Explorer in one click. See Preview Step Data.
Two new workflow steps — Directory Listing and Row Count Assert. Directory Listing writes a table of the files in a document directory (with an optional file pattern and a subdirectory option) so downstream steps can act on whatever files are present. Row Count Assert is a data-quality gate that fails the workflow when two tables have different row counts. See Directory Listing and Row Count Assert.

Changed

AI cost-tracing is more accurate and harder to mislead. What-if estimates are now real per-target numbers that add up across a fan-out, rather than upper bounds, and can be scoped to a single target — and pointed at a driver or basis table instead of the input pool, the agent declines the estimate and explains why. Driver-reweight what-ifs are exact: they move only the amount the driver actually splits and leave any pass-through cost untouched. The agent now recognizes and lowers its confidence on misleading shapes — a total whose large ups and downs offset to a small net change, or a change that is mostly pass-through rather than driver-governed — naming how much the driver-controlled amount actually moved and no longer overstating a single “dominant factor” when the split is unreliable. It also explains dimension-hierarchy allocations instead of leaving them unattributed. Data-freshness checks now cover the driver and basis tables, results produced by more than one allocation step can be traced by picking the step — or automatically when a filter narrows to rows only one step produced — and a year-over-year comparison with no prior-year data now says so rather than calling the change “normal”. See Tracing Allocations with an MCP-Connected AI Agent.
Alteryx import ends on a conversion report. Importing an Alteryx workflow now summarizes how many steps mapped with high confidence and how many need review, and lists the lower-confidence steps with notes so you can check them before running; each imported step also carries its confidence as a memo on the canvas. Packaged report macros now import as hand-editable Report steps, and a new guide plus a downloadable workflow batch-convert AMP-format .yxdb files to CSV so their data imports losslessly. See Migrate Alteryx Workflows and Convert .yxdb Files for Import.

Fixed

Join steps — Output Columns fixes. For Inner, Outer, Anti, and Cross Join, the Output Columns tab now lists only the columns you mapped from the two sources and drops stale rows when you remove a mapped source column; the column-mapper toolbar and multi-row selection are back; and Summarize is now a toggle that’s off by default, so a join no longer forces its output to be grouped — and when you turn it on, new columns default to an aggregation based on their data type, which you can override per column. Multi-Table Join output columns now follow renames and removals in the join-graph designer. See Table Inner Join and Multi-Table Join Step.
REST Request reliability fixes. Variables now substitute reliably in JSON request bodies, per-request timeout and retry settings are honored (retries limited to idempotent methods, so a POST/PUT/PATCH is never re-sent), rate-limited responses are retried with backoff, and request credentials are masked in any captured raw JSON. A test request now shows the response status code even on success, and a single-object response imports as one row. Sensor- and webhook-triggered workflows can also read the trigger’s metadata and inbound payload as variables. See REST Request Step.
Table Pivot column names keep their spaces again — a Table Pivot step names its new columns after your source category values exactly as they appear (for example, Freight Revenue), rather than replacing the spaces with underscores. A new Sanitize new column names option on the step turns the underscore form back on when you want it (spaces and punctuation become underscores, so Freight Revenue becomes Freight_Revenue); workflows imported from an Alteryx CrossTab tool enable it automatically. See Table Pivot.
AI cost-tracing fixes. Cost-tracing now handles model column names that contain spaces or symbols, resolves tables given by their friendly name, works on results from model-run steps and project copies, and works across all workspace types. Anomaly detection now flags unusually large swings and keeps quarter-labeled periods in real-time order across a year boundary, and two mislabeling bugs — a “from zero” jump when the weighting basis started empty, and a wrong-signed percentage change when the prior value was negative — are corrected.

Removed

Retired legacy import and export steps — the SAS7BDAT import, HTML import, and HTML export steps have been removed. Use the current import and export steps for these formats instead.