This article collects the errors and blocked states you can meet while building, testing, running, managing, or promoting a tool, with the cause of each and how to resolve it. If you've seen a message in Tool builder, the Run console, the Tools page, or an ILM pull and want to know what to do next, you're in the right place. Errors inside a lookup or import's own application call are covered by that connector's articles instead.
Run test is disabled
Symptom — Run test in the Tool builder header is grayed out, with the message Configure tool input and output to start a test run.
Cause — The tool input, the tool output, or both aren't configured.
Resolution — Open each step that shows Configure on the canvas and save it. See Configure tool input and Configure tool output.
Tool has no steps configured
Symptom — Selecting Run test shows Tool has no steps configured. Add at least one step before executing the tool.
Cause — The tool has an input and output but no lookup, import, or other logic step between them.
Resolution — Add at least one step. See Add lookups and imports to a tool.
Mock input is rejected
Symptom — The Mock tool input tab won't save, or Send on the Preview tab returns a validation error naming a property.
Cause — The mock input isn't valid JSON, or it doesn't satisfy the input schema (a required property is missing, or a value has the wrong type or fails a validation).
Resolution — Fix the JSON, or select Reset to sample data to regenerate it from the schema and edit again. See Configure tool input.
Tool output validation failed
Symptom — A step shows N errors after a test run, and Test run errors lists Tool output validation failed: must have required property '<name>' with the code tool_output_schema_validation_failed. The same error is returned to a caller at runtime.
Cause — The mapped result doesn't satisfy the output schema. Usually a required property isn't mapped, often because the output schema was changed after the mapping was built.
Resolution — Open the Tool output step's Mapping. If the mapper warns that the schema changed, select Auto-populate response fields, then map every required property. See Map data in a tool.
Test run was canceled
Symptom — The Run console shows the run as canceled, or a message appears while the run is in progress.
Cause — A message appears after 100 seconds to warn that the run is long. A run is canceled automatically after two minutes, or when a script throws an uncaught exception, or when you select Cancel test run.
Resolution — Reduce the work in the test (fewer records in the mock data, or mock response data instead of Populate with live data), and fix any script error shown in Test run errors. See Test a tool in Tool builder.
Unsupported keywords were removed from the schema
Symptom — After switching Create schema from JSON to Manual, a dialog lists keywords that will be removed, and the schema loses them.
Cause — Manual mode supports only the validations in Additional settings; keywords such as anyOf, oneOf, allOf, $ref, if/then/else, and uniqueItems exist only in JSON mode.
Resolution — Cancel the change and stay in JSON mode, or accept the removal if you no longer need those keywords. See Tool schema and validation reference.
An AI agent rejects the tool's schema
Symptom — An agent fails to load or call the tool, and the provider reports an unsupported schema keyword.
Cause — The model provider doesn't accept a keyword the schema uses. OpenAI strict mode rejects allOf, oneOf, and uniqueItems.
Resolution — Replace oneOf with anyOf, and restructure allOf and uniqueItems. See Tool schema and validation reference.
Tool run fails for a missing connection
Symptom — A flow, API, agent, or MCP server fails when it calls the tool, with an error about a connection that isn't configured.
Cause — The caller hasn't bound a connection for one of the connections the tool (or a nested tool) uses. Every connection must be supplied by the caller.
Resolution — Open the caller's tool step or tool settings and select a connection for each one listed. See Use a tool in flows, APIs, AI agents, and MCP servers.
Unable to delete tool
Symptom — Deleting shows Unable to delete tool. This tool is referenced by at least one resource. Before you can delete this tool, you must edit each resource to remove the reference.
Cause — A flow, API, AI agent, MCP server, or tool references the tool.
Resolution — Select Actions > Used by on the Tools page, remove the reference from each resource listed, and then delete the tool. See Delete a tool.
Proceed with caution banner
Symptom — A Proceed with caution banner appears when you open a tool.
Cause — Another resource references the tool, so changes you make affect that resource too.
Resolution — Select Used by in the banner to see what depends on the tool. To change the tool for one caller only, clone it and point that caller at the clone. See Clone a tool.
Tool shows Expired on the Tools page
Symptom — A tool's row shows Expired, and the tool can't be opened, cloned, moved, or deleted from the list.
Cause — The tool was created as a draft through the REST API and wasn’t saved within 48 hours. Draft tools expire after 48 hours and are removed automatically.
Resolution — Recreate the tool. An expired draft can’t be restored from the recycle bin.
No entries on the Logs tab
Symptom — The tool's Logs tab is empty although the tool is being called.
Cause — One of: the invoking agent or MCP server is set to No logging; execution logging isn't enabled for the account; the tool is called from a flow or API (those are logged with the flow or API); the invocations happened before logging was enabled; the date range doesn't cover them (the default is Last 5 mins); or the tool was deleted.
Resolution — Change the caller's log level, enable execution logging, widen the date range, or look in the flow run or API logs. See View tool invocation logs.
No debug entries for a step
Symptom — View debug logs shows nothing for a run you expected to see.
Cause — Debug logging captures only activity after you select Start debug, and stops when the duration you chose ends (Next 15 minutes by default). Real invocations appear under Debug logs and test runs under Test run debug logs.
Resolution — Start debug logging, then run the test. See Test a tool in Tool builder.
ILM promotion is blocked
Symptom — A pull or revert reports a validation error on a tool and doesn't run.
Cause — A connection the tool uses has no connection mapping at the destination, or a resource the tool depends on is missing.
Resolution — Add the connection mapping, or include the missing dependency in the snapshot, and retry. See Promote tools with ILM.