ioPipe

YOUR DATA. A CLEAR PATH FORWARD.

The next dataset.
Not the same
busywork.

Turn the steps between raw files and useful answers into a workflow you can see, check, and run again.

ioPipe brings data preparation, statistics, maps, and AI together on one visual desktop canvas.

Explore what’s possible See how it works

Built for analysts and researchers · macOS download coming soon

THE WORK IS VISIBLE

Every connection is a step you can inspect.

ioPipe canvas showing sample rows in the court rulings pipeline, with a quarantine branch for extraction errors and SQLite, Parquet, and CSV outputs
IOPIPE CANVAS · ROW PREVIEWScourt_rulings.iopipe

SAVE THE METHOD, NOT JUST THE OUTPUT

A clean dataset is useful.
A repeatable path to it is better.

When the next survey wave lands or another batch of files arrives, the work shouldn’t start over. In ioPipe, your preparation, analysis, and exports stay connected in a saved pipeline.

HOW IT WORKS

Build it. Check it. Run it again.

Work visually, with the detail close at hand.

01 Connect the steps
Bring in your files, then connect nodes to clean, join, and analyze them. Branch the same data into a table, a chart, or several export formats.
02 See what changed
Inspect sample rows at each step before running the full dataset. Route extraction errors into a separate output so they’re available for review.
03 Keep what works
Save the whole graph as an .iopipe file. Reuse it with new inputs, share the method, or run it from the command line.

ONE CONNECTED WORKFLOW

From “what’s in these files?”
to “here’s what we found.”

Keep the route from source to result in view.

↳ Make messy data usable
Extract HTML tables, clean text, handle missing values, and join datasets. Read survey formats including SPSS, SAS, and Stata.
↗ Follow through to the answer
Score survey scales, compare groups, and fit regression models. Turn the results into charts or maps in the same graph.
✳ Give AI a defined job
Classify reviews or extract fields with a configured local or hosted model. Cache responses and replay them as you refine downstream steps.
⇄ Take the work with you
Export datasets to CSV, Parquet, SQLite, and more. Save charts as SVG, PNG, or PDF. Share the pipeline alongside the result.

Built-in MCP support. Connect a compatible AI assistant through the Model Context Protocol (MCP) to build and edit pipelines, preview sample rows, and start runs.

REAL WORKFLOWS, CONCRETE OUTPUTS

Start with a question. End with something useful.

01 / DATA EXTRACTION

Turn court pages into a dataset

Read scraped HTML, extract ruling tables, rename the columns, and filter by year. The same cleaned data goes to CSV, Parquet, and SQLite. Pages without a matching table go to a quarantine file.

The sample run exported seven rulings and flagged one page with no matching table.

View the simplified pipeline diagram

Three rows from the CSV export
CaseDateOutcome
2019-442019-03-02Upheld
2019-452019-03-02Reversed
2024-022024-02-20Dismissed

02 / EXPERIMENT ANALYSIS

Find where a price-drop tag works

Aggregate weekly iced tea sales to the store level. Then use a two-way ANOVA to test whether the tag’s effect depends on shelf placement, and chart the six group means.

In the simulated data, end-cap displays averaged 69 units per week with a tag and 55 without. Eye-level displays averaged about 50 either way.

Grouped bars comparing weekly sales with and without a price-drop tag at three shelf placements
Interaction chart exported by ioPipe

03 / GEOSPATIAL ANALYSIS

Map where the 2024 polls missed

Aggregate archived 538 polls and county election returns to the state level. Join the results to Census state boundaries and map the gap between the final poll average and the result.

All ten states with at least ten late polls shifted toward Trump relative to their poll averages in this run.

US choropleth showing the difference between the 2024 election result and final state poll averages
Polling error map exported by ioPipe
More example projects
  • Brand tracking — combine survey waves, score scales, and export a codebook and SPSS file.
  • Satisfaction drivers — use factor analysis and regression to study satisfaction and score churn risk.
  • Campaign test — validate a send log, join CRM records, and compare email offers by segment.
  • App redesign — test how an app redesign affects spending and whether the indirect effects vary by age.
  • Customer reviews — extract review fields from HTML and compare human and model topic labels. AI steps require a configured model.
  • Product reviews — classify real reviews by sentiment using a local model adapter. This example requires local model setup.
  • Store catchment — find the nearest store, group customers by distance, and map areas with weak coverage.

UNDER THE HOOD

Simple to follow. Plenty to work with.

Explore the built-in tools for preparation, analysis, visualization, and export.

Explore the full node catalog, MCP, and Python options

Built-in node catalog

Expand a category to see its nodes. Each entry shows the name from the app and what the node does.

File discovery and input readers 5
  • Files Find files in a folder using filename patterns.
  • Read CSV Combine CSV or TSV files with matching headers into one table.
  • Read SAS Import SAS data and transport files with variable labels and formats.
  • Read SPSS Import SPSS data with value labels, missing-value codes, and case weights.
  • Read Stata Import Stata data with variable labels, value labels, and date formats.
HTML parsing and extraction 3
  • Extract Fields Pull text or attributes from HTML using CSS selectors.
  • Extract Table Turn HTML tables into rows and match columns by their headers.
  • Parse HTML Decode HTML pages using their detected character encoding.
Column transformations 18
  • Automatic Recode Assign numeric codes to distinct values and keep the originals as labels.
  • Clean Text Trim whitespace, strip HTML, and treat placeholders like N/A as missing.
  • Compute Variable Create or update a column using a SQL expression.
  • Count Values within Cases Count how many selected columns match a value or range in each row.
  • Date & Time Parse dates, extract date parts, and calculate or add time intervals.
  • Dummy Coding Create a binary indicator column for each category.
  • Lag Copy values from earlier rows, optionally within each group.
  • Lookup / Map Values Replace values using a lookup table, such as a codebook.
  • Pseudonymise Hash IDs, remove direct identifiers, or reduce the precision of identifying fields.
  • Recode Replace individual values or ranges according to rules you define.
  • Rename Rename columns individually or with a pattern.
  • Replace Missing Values Fill gaps with a mean, median, fixed value, or last valid value.
  • Reverse Score Reverse the direction of a rating scale while preserving missing values.
  • Scale Score Sum or average survey items that meet a minimum response count.
  • Select / Drop / Reorder Choose which columns to keep and how to order them.
  • Standardize Rescale values using z-scores, min–max scaling, or the median and interquartile range.
  • String Functions Change case, replace text, or extract and combine parts of strings.
  • Visual Binning Group numeric values into labeled intervals.
Row operations 7
  • Add Row Number Number rows across the table or within each group.
  • Explode Turn each item in a list or delimited field into its own row.
  • Identify Duplicate Cases Flag rows with duplicate keys or keep one row per key.
  • Limit / Range Keep a fixed number of rows, with the option to skip rows first.
  • Sample Draw a repeatable random sample by fraction or exact row count.
  • Select Cases Keep or flag rows that match a SQL condition.
  • Sort Cases Order rows by one or more columns while preserving the order of ties.
Joins, aggregation, and reshaping 7
  • Aggregate Calculate group totals, averages, counts, and other summaries.
  • Join Combine two tables by matching key columns.
  • Merge: Add Cases Stack two tables, matching columns by name.
  • Merge: Add Variables Bring columns from a lookup or matching table into the original rows.
  • Restructure: Cases to Variables Pivot repeated rows into separate columns for each ID.
  • Restructure: Variables to Cases Turn groups of columns into rows and add an index to identify each group.
  • SQL Run a SQL query across up to three input tables.
Labels, weights, and validation 7
  • Apply Labels Swap coded values for their labels or convert labels back to codes.
  • Define Missing Values Mark specific values or ranges as missing for later steps.
  • Split File Tell later steps to calculate results separately for each group.
  • Validate Data Check data against rules and flag, report, or drop rows that fail.
  • Value Labels Attach readable labels to coded values, such as survey responses.
  • Variable Properties Set column labels, measurement levels, roles, and display formats.
  • Weight Cases Choose a column that determines how much each row counts in later analyses.
Data profiling and summaries 7
  • Crosstabs Compare two categorical variables using counts, percentages, and tests of association.
  • Data Profile Review each column’s type, missing values, range, and common values.
  • Descriptives Summarize numeric columns with counts, averages, spread, and range.
  • Explore / Quantiles Explore numeric distributions with medians, percentiles, and interquartile ranges.
  • Frequencies Count how often each value appears and report percentages and missing responses.
  • Means by Group Compare averages, standard deviations, and counts across groups.
  • Missing Value Patterns Find which columns have missing values and which tend to be missing together.
File, database, and report outputs 10
  • Codebook Export a Markdown or HTML reference for variables, labels, and missing-value codes.
  • Export Chart Save a chart as SVG, PNG, PDF, or Vega-Lite JSON at a specified size.
  • Results Table Export an APA-style results table to Word, LaTeX, HTML, or Markdown.
  • Save Model Save a fitted model for sharing or prediction in another pipeline.
  • Write CSV Export rows to a comma- or tab-separated file.
  • Write JSONL Export one JSON object per row.
  • Write Parquet Export typed columns and variable metadata to a Parquet file.
  • Write SPSS Export an SPSS file with labels, missing-value codes, and case weights.
  • Write SQLite Create, replace, or append to a table in a SQLite database.
  • Write Stata Export a Stata data file with variable labels, value labels, and formats.
Statistical tests and models 12
  • ANOVA Compare group means across factors, test interactions, and account for covariates.
  • Chi-square Test whether observed counts fit an expected distribution or whether categories are associated.
  • Correlation Matrix Measure associations with Pearson, Spearman, Kendall, or partial correlations.
  • Linear Regression Model a numeric outcome and estimate how predictors relate to it.
  • Logistic Regression Model a binary outcome and estimate probabilities and odds ratios.
  • Mediation & Moderation (PROCESS) Test indirect effects and how effects vary across conditions using PROCESS models.
  • Multiple Comparisons Adjust p-values to account for multiple tests.
  • Nonparametric Tests Compare groups or paired measurements using rank-based or sign tests.
  • PCA / Factor Analysis Reduce a set of variables to components or estimate underlying factors.
  • Predict Apply a fitted model to new rows to generate predictions or factor scores.
  • Reliability Assess how consistently survey items measure a scale using alpha and omega.
  • t-tests Test a mean against a benchmark or compare independent or paired means.
Charts and result tables 12
  • Bar Chart Compare category totals, percentages, or averages with grouped or stacked bars.
  • Box Plot Show medians, quartiles, and outliers across groups.
  • Chart Style Apply consistent colors, fonts, labels, and sizing to charts.
  • Coefficient Plot Plot regression estimates and confidence intervals side by side.
  • Heatmap Show a correlation matrix or summary table as a grid of colors.
  • Histogram Show how numeric values are distributed across intervals.
  • Line / Time Series Plot trends across dates or numeric intervals, optionally with a moving average.
  • Map Make a map with up to three layers and color features by value or category.
  • Model Diagnostics Plot residuals and leverage to check a linear regression fit.
  • Q-Q Plot Compare observed and normal quantiles to check whether data follow a normal distribution.
  • Scatter Plot the relationship between two numeric columns, with optional groups or a fitted trend.
  • Small Multiples Arrange several charts in one figure.
Geospatial processing and maps 34
  • Aggregate & Suppress Summarize points by area and hide results where too few records contribute.
  • Aggregate to Areas Count points and summarize their values within polygons or H3 cells.
  • Areal Interpolation Transfer values between boundary systems based on how much their areas overlap.
  • Buffer Expand or shrink features by a specified distance.
  • Clip Trim lines and polygons to a boundary and discard points outside it.
  • Dissolve Merge geometries by group and summarize their attributes.
  • Distance Measure the distance between two geometries or to a fixed location.
  • Explode Multi-part Split a multipart geometry into one row per part.
  • Gazetteer Match Match place names in fields or text to an offline directory of locations.
  • Geocode Turn addresses or place names into coordinates and report match scores.
  • Geomask Shift points by a repeatable random distance before sharing location data.
  • Geometry to Columns Convert shapes to coordinates, bounding boxes, or standard text formats.
  • Grid / Tessellate Cover an area with square or hexagonal cells for mapping and analysis.
  • H3 Index Assign points or polygons to cells in the H3 hexagonal grid.
  • Make Points Build point geometries from latitude and longitude or projected coordinates.
  • Measure Calculate area, length, perimeter, centroids, or other geometry measurements.
  • Moran's I Test whether nearby places tend to have similar values.
  • Nearest Neighbour Find the closest features in another layer and report their distances.
  • Parse Geometry Read geometry from formats such as WKT, WKB, GeoJSON, or H3 cell IDs.
  • Read GeoJSON Import GeoJSON features with their geometry and attributes.
  • Read GeoPackage Load a feature layer and its coordinate system from a GeoPackage.
  • Read GeoParquet Load geometry and attributes from GeoParquet, optionally within a bounding box.
  • Read Shapefile Import geometry and attributes from plain or zipped shapefiles.
  • Reproject Convert geometry to a different coordinate reference system.
  • Simplify Reduce the number of vertices in lines and polygons for smaller files and faster maps.
  • Spatial Filter Keep features that lie within or intersect a chosen area.
  • Spatial Join Attach attributes from another layer based on where features intersect or how close they are.
  • Spatial Lag Calculate the weighted average of neighboring features’ values.
  • Spatial Weights Define which features are neighbors and how much each neighbor counts.
  • Validate & Repair Geometry Check for geometry problems and repair, flag, or drop invalid features.
  • Write GeoJSON Export features and attributes as a GeoJSON collection or sequence.
  • Write GeoPackage Save a feature layer with its coordinate system and spatial index in a GeoPackage.
  • Write GeoParquet Export features to Parquet with standardized geometry metadata.
  • Write Shapefile Export a feature layer as an ESRI shapefile and its companion files.
AI models and coding agreement 5
  • Agreement Measure how often two coders agree and where their labels differ.
  • Embed Convert text to numeric vectors for similarity search or clustering.
  • LLM Classify Assign one or more predefined categories to each row with a language model.
  • LLM Extract Use a language model to extract structured fields from unstructured text.
  • LLM Prompt Send a customized prompt for each row and store the response as text or JSON.
Reusable graph inputs and outputs 2
  • Library Input Bring data into a reusable group of nodes through a named port.
  • Library Output Expose a result from a reusable group of nodes through a named port.

Model Context Protocol (MCP)

In the desktop app, open File → Connect an AI Assistant to start a local MCP server and copy its connection settings. Command-line clients can launch the server with iopipe mcp --workspace DIR.

Optional Python sidecar

Use Python reference implementations for Descriptives, Means, Frequencies, and Crosstabs during full runs. Enable it in the desktop app under Settings → Python sidecar and turn it on from the toolbar, or point the command line at your configured sidecar with iopipe run FILE.iopipe --sidecar verify.

Previews use the built-in implementation. During a full run, ioPipe falls back to the built-in node when the Python implementation declines a configuration. The node’s run statistics show which implementation ran.

BUILD ON YOUR OWN WORK

Make the next run
the easy part.

ioPipe for macOS is on its way. Explore a sample workflow while the public download is being prepared.

Explore the examples macOS download coming soon