Tools

Tools are actions. The agent calls them to create, run, and inspect tests, and to drive the guided flows.

Every tool except authenticate requires a signed-in Cycle session. See Signing in to Cycle.

Authoring and execution

Tool	What it does
authenticate	Opens a browser login window so you can sign in to Cycle. This is the only tool that works while you are logged out. The agent calls it when a request returns an authentication error, or you can ask for it directly: "Authenticate with Cycle."
create_project	Scaffolds a new Cycle project directory tree, writes the .cycproj, and writes a user settings file. Useful when starting from scratch in an empty folder.
validate_feature	Parses a .feature file and returns syntax errors with line numbers. Does not execute the file. The agent should call this every time it edits a Feature File, to catch errors before runtime.
run_test	Executes a .feature file with cycle-cli. Returns the exit code, truncated output, and the path to the directory containing the results .db file. That directory is what you pass to the cycle://run-errors/{output_dir} resource to inspect failures.
run_step	Executes one step from any installed plugin (WebDriver, File-System, and so on) in a persistent live session. The browser session survives across calls, so the agent can open Chrome, navigate, click, and take a page snapshot to discover real element locators before writing them into a .feature file. This is how the agent avoids guessing at selectors. See Intelligent App Exploration for the workflow built on it.
reset_step_session	Ends the current run_step scenario and starts a clean one, for when the live session gets stuck. It disposes the browser session, so the agent must reopen one with WebDriver.startLocal, but it leaves the plugin and engine processes running so recovery skips the slow startup. It does not revive a crashed plugin.
assign_variables	Resolves $variable placeholders in a Scenario against Dynamic Data sources, and reports which names could not be resolved. Useful for checking a test's data wiring before a run.
scenario_search	Finds reusable utility Scenarios in a project's Utilities tree, ranked by how well they match what the agent is trying to do. This is how the agent reuses an existing Scenario instead of authoring a new flow.
wms_validate_dynamic_tokens	Checks that the ?token placeholders in a CSV resolve against the project's Dynamic Data inventory, catching a broken data reference before a run.

Blueprints

These tools author and run the test data a test depends on. See Intelligent Data for the workflow, and Blueprints for the file format. The main agent usually delegates this whole loop to the blueprint-author subagent.

Tool	What it does
blueprint_validate	Structurally checks a Blueprint document without touching any system. Returns either a pass or the list of findings to fix. The agent should call this after every edit, before running anything.
blueprint_run	Runs a Blueprint end to end, creating its records and returning every declared field value plus a verdict. Takes exactly one of a catalog id, a file path, or inline YAML. Can also run offline against a local SQLite database when no live system is available.
blueprint_teardown	Reverses a Blueprint run, removing what it created in reverse dependency order. This is a two-call tool: the first call previews what would be removed and deletes nothing, and only a confirmed second call executes.

Blueprint discovery is plain file reading

There are no tools for browsing a blueprint library. The library is a folder of ordinary YAML files, so the agent reads blueprint-catalog.yml, a system configuration, and individual Blueprints directly. The three tools above are the only Blueprint tools.

Discovery

These tools drive the discovery questionnaire and test-plan generation. You do not invoke them individually. The begin-discovery prompt drives all of them for you, which is the documented way to use Discovery. See Discovery.

Tool	What it does
discovery_engine	Starts or resumes Discovery. Determines which Cycle project to work in, prepares the questionnaire on first entry, and returns the input-source hub with each source's status and whether a test plan can be generated.
discovery_get-next-questions	Lists the currently applicable, unanswered questions, grouped by phase and area, along with a summary of how many remain per section.
discovery_answer-question	Records one answer, and clears any answers whose dependencies the change invalidated.
discovery_review-questions	Reports the currently applicable questions, answered and unanswered, with each one's current value and who supplied it. Read-only, and the basis for revisiting a decision.
discovery_get-status	Reports completion counts and percentages per area, per phase, and overall, counted over applicable questions only.
discovery_generate-test-plan	Turns the recorded answers into a recommended test plan and a playlist. Requires the questionnaire to be complete, and reports conflicts rather than overwriting files that differ.

Formulation

These tools turn test specifications into Cycle tests. The generate-tests-from-specs prompt orchestrates them.

Tool	What it does
formulation_list-bundle-members	Lists the test specifications and reports in a Test-Specs folder, with a readiness category for each specification so the set can be triaged.
formulation_read-bundle-member	Reads one specification as a structured model, or one report as text.
formulation_generate-tests-from-specs	Plans how to generate tests from selected specifications, one test per specification. Produces an authoring brief for each and skips specifications an existing test already covers. It does not author the tests itself.
formulation_validate-generated-test	Runs one generated .feature through parse, step resolution, binding, and execution, and returns a staged verdict with the failing step and a failure classification.

Blue Yonder

Off by default

These tools are available only when Blue Yonder context is enabled on the server, which is off by default. If your agent does not list them, that is why. Enable them by setting CYCLE_BLUE_YONDER_CONTEXT_ENABLED=true in the .env file

Tool	What it does
classify_failure	Classifies a failed run into a category (popup, locator, environment, data, or timing) and returns a ranked recovery suggestion.
agent_web_selectors	Suggests Blue Yonder web element selectors for an intended screen or flow, drawn from a curated locator inventory.
moca_exec	Reads Blue Yonder WMS datastore state by running a MOCA command, for checking inventory, transaction status, or reference data while authoring or debugging a test. The connection is configured on the server, so credentials never appear in a tool call.
wms_autofix_inventory	Corrects known Blue Yonder Inventory grid and search anti-patterns in a .feature file. Advisory: it returns the corrected text rather than writing the file.
formulation_moca_finalize-tests-live	Finalizes a batch of converted tests against the live Blue Yonder WMS and reports validated coverage per specification. Resumable across sessions.