Skip to content

Migrate KNIME Workflows

PlaidCloud is a drop-in replacement for KNIME. 98.5% of KNIME node configurations convert fully, and 205 of the 207 core node types are Full: each becomes standard PlaidCloud workflow steps that do the node’s work. The converter turns a KNIME workflow archive into runnable Advanced workflows, keeps the original graph, and returns a conversion report that names anything you need to set. The KNIME Conversion Matrix lists every node.

Conversion reads only the archive you point it at. It never contacts the system a KNIME node reads from or writes to.

Conversion is a one-time step. Converting the same archive again creates new workflows, and edits you made in PlaidCloud are never merged with a later export.

  • One Advanced workflow for each KNIME workflow. A .knwf gives one workflow, and a .knar gives one for every workflow in the group.
  • Each KNIME node becomes the matching workflow step, project variable, or canvas object, with upstream and downstream relationships preserved. Metanodes and components that hold steps arrive as collapsed groups, and their nodes run as ordinary steps. A metanode or component nested inside another becomes a group of its own, named by its path, such as “Outer › Inner”.
  • Annotations on the top level of the workflow become notes on the canvas.
  • The workflow opens already tidied: steps are laid out cleanly, groups from the source are kept, and notes are moved to just above the flow, keeping their arrangement. This happens the first time someone with edit access opens it. After that, your own layout is saved and never rearranged. If the workflow or project is locked, you see the tidied layout but it isn’t saved.
  • Flow variables, and Configuration and Widget nodes, become project variables holding their defaults. Switches and loops whose values are known when you convert are pruned and expanded into the steps they run.
  • The KNIME expression languages — Expression, String Manipulation, Math Formula, Column Expressions, and Rule Engine — convert to PlaidCloud expressions.
  • Data files that ship inside the archive are uploaded to a knime/ folder in the project’s default Document account, one folder per workflow, and the converted readers import them from there.
  • Views, charts, and color managers have no data effect. They are dropped and listed in the conversion report.
  1. Export the workflow from KNIME Analytics Platform as a .knwf file. To convert several workflows at once, export their workflow group as a .knar file. Keep the files your readers use in the workflow’s data folder so they ship inside the archive.
  2. Upload the archive to a Document account you can read. In the Analyze app, Browse on the import window uploads it from your computer.
  3. Choose the PlaidCloud project that should receive the workflows. You must be an Architect or Manager of the project, and the project must have a default Document account.
  4. Note the PlaidCloud tables, files, and connections you want the workflow’s readers and connector nodes to use. See Bind Sources.

Convert from the Analyze app, the REST API or an MCP-connected AI agent. The REST API and the MCP tool take the same fields; the window takes the ones it shows.

  1. Open Analyze and choose Tools → Import KNIME Workflow.
  2. Under Import Source and Target, choose the Target Project and the KNIME Archive (.knwf or .knar) in a Document account. Select Browse to upload the archive from your computer.
  3. Under Object Prefixes, optionally set the Workflow Prefix, Step Prefix and Table Prefix. The table prefix defaults to the workflow’s name.
  4. Under Source Bindings (optional), optionally bind a reader to an existing table in the target project: select Add, then enter the KNIME Reader (its node id, such as 12:3, its node name or its file name) and the Table ID. To bind connections, documents or tables in other projects, use the REST API or the MCP tool. See Bind Sources.
  5. Select Import.

The window shows progress while it converts the archive, then the conversion report: step counts (Steps, High confidence, Medium, Low and Not converted), the nodes that could not be converted and the steps downstream of them, partial conversions, warnings, the source files to upload, and the nodes with no step. A conversion with nothing to review says so and the window closes on its own.

Field Required What it does
source_account_id Yes Document account that holds the KNIME archive.
source_file_path Yes Path of the .knwf or .knar file within that account.
project_id Yes Project where the workflows are created.
workflow_prefix No Prefix added to workflow names.
step_prefix No Prefix added to step names.
table_prefix No Prefix added to table names. Defaults to the workflow’s own name. A prefix that would reuse an existing table name is refused before anything is created. For a .knar, each workflow’s prefix is followed by that workflow’s name.
source_bindings No Binds readers and connector nodes to data and connections you already have, up to 500 entries. See Bind Sources.

Send the fields as a JSON body to:

POST /rest/v1/analyze/knime/convert
{
"source_account_id": "<document account id>",
"source_file_path": "/knime/monthly-close.knwf",
"project_id": "<project id>"
}

An MCP-connected AI agent calls the knime_convert tool with the same fields. See AI Agents (MCP) to connect an agent. The tool changes your project, so review the call before you approve it.

Use source_bindings to point a KNIME reader at data that already exists in PlaidCloud instead of the copy in the archive. A key is a node id (such as 12:3), a node name, a bundled file path (such as data/sales.csv), or a file name; the first match in that order wins. A value is one of:

  • An analyzetable_ id for a table in the target project. The table replaces the reader.
  • {"project": "<project id>", "table": "<table id>"} for a table in another project you can read. The reader becomes an Import Project Table step that copies it. The source project must share Read with the target project, or the step warns. A table governed by row access is refused.
  • {"account": "<document account id>", "path": "/finance/budget.xlsx"} for a document you can read. The reader imports it.
  • {"connection": "<connection id>"} for a database or REST connector node. The converted steps run on that connection.

The Analyze app window binds readers to tables in the target project by table id. Bind connections, documents and tables in other projects through the REST API or the MCP tool.

"source_bindings": {
"Sales": "analyzetable_1234",
"data/budget.xlsx": { "account": "<document account id>", "path": "/finance/budget.xlsx" }
}

A bound table that does not exist, or a key that matches no reader, is a warning. The bindings applied, with their project and table names, are listed in the conversion report under bindings.

A reader whose file is not in the archive reads from inputs/<file name> in the workflow’s folder of the default Document account, and the conversion report lists the path under missing_inputs with a warning. Upload the file there and run the workflow. Until the file is there, the import finishes with a warning and creates an empty table, and the workflow continues.

KNIME credentials are never migrated. Each step that signs in to another system reads its credentials from a PlaidCloud connection or a project variable you set, and the step’s warning in the conversion report says which.

KNIME What You Set
Database connectors A database connection: bind it with {"connection": "<id>"} or pick it on the step.
REST and web service nodes that sign in A REST connection: bind it with {"connection": "<id>"} or pick it on the step.
Speech Recognizer (IBM Watson) A REST connection whose base URL is the Watson service URL the warning names, with the user apikey and your API key as its Basic login. Bind it to the node.
Email Connector, with Email Reader or Email Mover A project variable holding the mail server password, with the name the warning gives. The host and user come from the Email Connector.
Excel Reader or Excel Writer with a workbook password A project variable holding the password, with the name the warning gives.
LLM connectors and authenticators An LLM connection on the AI step.

See Manage Workflow Variables to set project variables.

The Analyze app window shows the conversion report when the import finishes, and the REST and MCP responses carry it. Read it before you run the workflows.

  • summary counts the steps created and how many mapped with high, medium, or low confidence.
  • warnings lists each warning with its node, step, and the action to take, including every credential or connection to set.
  • missing_inputs lists the Document paths to upload files to for unbound inputs.
  • bindings lists the bindings applied.
  • table_prefix is the prefix used. For a .knar, workflows lists each workflow with its own prefix.
  • uploaded_files, beside the report, lists the bundled files uploaded to the default Document account.
  • dropped_tools lists the views and charts that were dropped.

A node without a conversion path stays on the canvas as a named placeholder. The placeholder passes its data through and warns with the KNIME node’s name, and every step downstream of it is flagged in the report, so a missing result is never silent. Rebuild the capability with a standard step when its result matters.

  1. Open the new workflow in the project.
  2. Set the project variables and connections the warnings name.
  3. Upload files for unbound inputs to the paths under Source files to upload in the window, or in missing_inputs.
  4. Run the workflow and review each step’s output table.

These are the only known differences from KNIME.

  • Crosstab. Converts to Table: Pivot. A few settings combinations give results that differ from KNIME.
  • Loop End (Column Append). Converts to Table: Inner Join. A few settings combinations give results that differ from KNIME.
  • Random draws. Row Sampler, Table Partitioner, and Equal Size Sampling make KNIME’s own random draw with Table: KNIME Random Sample, so a node with a seed selects the same rows as in KNIME. The draw takes a table of at most 10,000,000 rows; a larger table fails the step.