Use this reference to find a request you can copy, a command for a known step, or the next check for an error. Jump to the section you need and expand the relevant group.

Use the request library for natural-language tasks, the command map for direct CLI work, and troubleshooting when a command or workflow does not behave as expected.
Figure 1. Choose the request library, command map, or troubleshooting path for the task at hand.

Request library

Copy, swap in your own workflow, app, or Business Object names, and send to your coding assistant after the guardrail request from Part 4.

Orient — show / hide

Try it: Tour the project

Inspect this AI Studio project. Tell me whether it uses the legacy src layout or app packages, list every artifact by type, and point out any workflow that hasn’t been validated or has no tests. Don’t change anything.

Try it: Explain a workflow

Explain what src/workflows/position_to_job_requisition.wf does in business terms: its routes, the Business Object functions it calls, and what each path returns. Don’t change anything.

Build — show / hide

Try it: Turn a use case into a build request

Use the aistudio skill to write a prompt I can give a coding assistant to build a standalone workflow for this use case: [describe users, intents, data, and expected answers]. Include Business Object identification, a suggested topology, wiring constraints, and finish criteria. Don’t build anything yet.

Try it: Find a Business Object

Search for an existing Business Object that can list, create, and update [records]. Show its functions and parameters, tell me whether it covers every operation I need, and recommend reuse or a new BO. Don’t create anything until I confirm.

Try it: Start an Agentic App

I want an Agentic App for [user] to [goal]. Run app intake and present the pre-build checkpoint: data source, app goal and user, first-load experience, query behavior, priority actions, communications, side effects, launch context, and validation scenario. Wait for my confirmation before creating any artifact.

Change — show / hide

Try it: Add a node safely

In [workflow file], add a condition guard before [node] so empty results return “[message]” instead of calling the next Business Object. Use file-backed patches for the expression, prettify, validate, and run the workflow test sync. Keep changes local.

Try it: Preview a change

Show me the exact file diff for adding a code node after FIND_POSITIONS in position_to_job_requisition.wf that counts the returned positions, using a dry run. Don’t write the change.

Apps — show / hide

Try it: Add a panel

Add a “[Panel heading]” panel to [app file], backed by [workflow file] and [BO file]. Data source: [query]. First load: [what it shows]. Query behavior: answer only from that result set. Priority actions: omit for now. Communications: omit for now. Validate the workflow and app and sync both test suites.

Try it: Inspect panel layout

Show me the panel layout of [app file]: each panel, the workflow agent behind it, its actions, and its communications. Don’t change anything.

Server — show / hide

Try it: Fetch for review

Fetch the latest DRAFT of app [APP_CODE] and every artifact it references, read-only. Validate the local files and give me an inventory with the status used for each, any gaps, and confirmation that nothing on the server changed.

Try it: Save a reviewed DRAFT

Save src/workflows/position_to_job_requisition.wf to its remote DRAFT. If the remote DRAFT is newer than my local file, stop and show me the version instead of overwriting it.

Debug — show / hide

Try it: Why did it take that path?

Run [workflow file] with “[message]” through the debug endpoint. Show me which nodes ran, the output of [node], and the first place the run diverged from what I expected. Don’t change the workflow.

Try it: Pin and retry

Pin this output on FIND_POSITIONS in the debugger: [paste JSON]. Run the workflow again with the same message and tell me what SELECT_POSITION produced. Clear the pinned output afterward.

Release — show / hide

Try it: Prepare a merge request

Summarize every artifact and test file changed on this branch, the validation results, and the test report links, in a form I can paste into a merge request. Confirm env.properties and test-reports aren’t staged.

Command map

The bundled CLI has more than 400 commands. You rarely type them; the assistant chooses them. These are the families you’ll see most, so you can follow what it’s doing. Run aistudio --help or aistudio <command> --help for the full list and options.

How every command runs

node .agents/skills/aistudio/scripts/aistudio.js <command> [options] from the project root. Don’t run commands through npm run or npm test; those scripts are for CI pipelines.

When to run a command yourself

Most of the time your coding assistant runs the commands. A simple rule for when to run one yourself: use the CLI directly when the next step is known and mechanical, and bring in an assistant when the work needs planning, interpretation, or changes across several files.

Run the CLI yourself

The next step is known, the command is mechanical, and you can supply the options with confidence.

  • Check the version, who you’re signed in as, or a command’s help
  • Validate a file you know
  • List or fetch server objects you can name
  • Run a focused workflow or app test suite
  • Save or publish artifacts you’ve already reviewed

Ask a coding assistant

The work needs planning, interpretation, multi-file changes, or careful sequencing.

  • Create or redesign a workflow from a description or a spec
  • Add nodes, panels, actions, templates, or tests
  • Fix validation or test failures across several artifacts
  • Decide which artifacts are needed and in what order
  • Coordinate the workflows behind one application

Make aistudio a command

The CLI is a script inside the skill, so the full command is node .agents/skills/aistudio/scripts/aistudio.js <command>. To type just aistudio <command>, add a shortcut to your shell profile once. The path is relative, so run it from your project root as usual.

macOS or Linux · add to ~/.zshrc or ~/.bashrc

alias aistudio='node .agents/skills/aistudio/scripts/aistudio.js'

Windows PowerShell · add to $PROFILE

function aistudio { node .agents/skills/aistudio/scripts/aistudio.js @args }

Then, from the project root

aistudio whoami
aistudio validate-workflow --file src/workflows/position_to_job_requisition.wf

Your coding assistant still uses the full path

The shortcut is for you. The skill always calls the script by its path from the project root, so you’ll see the long form in the assistant’s output. The tables below use the short aistudio form.

Session and project — show / hide
Command What it does
authenticate · whoami · logout Browser sign-in, check the signed-in user, remove stored credentials.
login-with-key Non-interactive sign-in with an OAuth key, for pipelines.
version · get-server-compatibility CLI version; server release and which skill operations it supports.
init · init-app-package Scaffold a blank legacy project, or a source-controlled app package.
Workflows — show / hide
Command What it does
list-workflow-families · list-workflow-products Valid family and product values for a new workflow.
do-create-workflow Creates a local .wf file.
do-create-node · do-modify-node · do-delete-node Adds, patches, or removes a node.
do-modify-node-edges Adds or removes edges between nodes by code.
do-modify-workflow-metadata · do-modify-workflow-variables Workflow-level metadata and workflow-scoped variables.
get-workflow-node-structure · get-nodes-metadata-by-code The graph by node code; node configuration summaries.
do-prettify-workflow · validate-workflow Tidies node layout after structural edits; validates and prints Test Impact.
do-set-workflow-default-model · do-set-workflow-node-model Sets a model from the current environment’s catalog.
do-fetch-workflow · do-save-workflow Pulls a workflow from the server; saves the local file to the remote DRAFT.
Agentic Apps — show / hide
Command What it does
do-create-app · validate-app Creates a local app file; validates it.
do-add-agent · do-modify-agent-config · do-remove-agent Places, configures, or removes a workflow agent in the app.
do-add-action · do-add-communication · do-add-template Adds actions, communications, and templates, each with modify and remove partners.
get-panel-metadata · list-configured-workflow-agents Panel layout and agent placement; agents configured in the app.
do-modify-page-pattern · do-modify-app-config Page layout and top-level app configuration.
do-fetch-app · do-save-app Pulls an app from the server; saves application DRAFT metadata.
Business Objects — show / hide
Command What it does
search-business-objects Recommends existing Business Objects for what you describe.
get-business-object-functions · get-bo-function-output-specification A BO’s functions; a workflow-ready output specification for one function.
do-create-bo · do-create-bo-function-from-operation Creates a local .bo; adds a function from an OpenAPI operation.
do-add-bo-function · do-add-bo-function-parameter · do-add-bo-function-example Adds functions, parameters, and examples by hand, each with modify and delete partners.
validate-bo · do-fetch-bo · do-save-bo Validate, pull, and save.
Agents, tools, topics, and deeplinks — show / hide
Command What it does
do-create-agent · do-add-agent-tool · do-add-agent-topic Creates an agent and attaches tools and topics.
do-create-tool --tool-type <type> Creates a Business Object, Deep Link, Document, Email, External REST, MCP, or Connector tool.
do-create-topic · do-add-topic-instruction Creates a topic and adds instructions.
do-create-deeplink · do-modify-deeplink-parameter Creates a deeplink and edits its URL-token parameters.
search-agents · search-workflows · search-external-rest-tools Finds reusable agents, workflows, and tools before you build new ones.
Other artifact types — show / hide
Family Typical commands
Approval processes do-create-approval-process · do-create-approval-rule · validate-approval-process
Policy Store and templates do-create-policy · do-generate-policy-functions · run-policy-test-cases · do-publish-policy
Document schemas do-create-document-schema · do-generate-document-schema · validate-document-schema
Function templates do-create-function · do-generate-function-implementation · run-function-test-cases
Connectors do-generate-connector-definition · do-create-connector-instance · validate-connector-instance
Ask Oracle plugins do-create-ask-oracle-plugin · do-add-linked-application · validate-ask-oracle-plugin
Run and debug — show / hide
Command What it does
run-workflow · run-workflow-chat Ad hoc runs through the debug or chat endpoint. Not tests.
get-nodes-executed-on-last-debug · get-debugger-results-for-nodes The path of the last debug run; what selected nodes received and returned.
do-node-override-output · do-modify-node-overrides Pin a node’s output, or change its configuration for debugging only. Both have clear partners.
list-pinned-outputs · get-debugger-node-override-schema · get-logs Pinned outputs, editable override keys per node type, and orchestrator logs.
Testing (ATLAS) — show / hide
Command What it does
get-workflow-test-sync-plan · get-app-test-sync-plan The read-only coverage plan for a workflow or app.
run-workflow-tests · run-app-tests · run-tests Run a workflow suite, an app’s panel tests, or the whole package gate.
get-workflow-test-final-summary · get-app-test-final-summary The evidence behind Validation and Insights.

The full testing command set will be covered in the upcoming ATLAS learning path.

Flags you’ll see most — show / hide
Flag Meaning
--file <path> The local artifact file a command works on.
--dry-run Show the exact file diff without writing it.
--app-package <name> · --modulename <name> Target package and module in an app-package repo.
--status DRAFT|PUBLISHED Which server version a fetch reads. Defaults to DRAFT.
--force Overwrite on a fetch or save conflict. Only with your explicit approval.
--inputs-patch @<file> File-backed JSON for expression-bearing node edits, so the shell can’t mangle quotes or {{...}}.
--verbose Redacted request and response diagnostics, for network troubleshooting.

Troubleshooting

What you’re most likely to run into, and what to do about it.

Four troubleshooting checks: project root and script path, authentication and credential access, changed expressions and node outputs, and local versus remote versions.
Pick the next check. Expand the matching symptom below for its specific guidance.
“aistudio: command not found” — show / hide

There’s no global executable. Run the bundled script from the project root: node .agents/skills/aistudio/scripts/aistudio.js <command>. If the skill is installed somewhere else, use that folder’s scripts/aistudio.js, still from the project root.

The command returns AUTH_REQUIRED — show / hide

The CLI needs you to sign in again. In a real terminal, run the interactive command it returns, then retry. If your assistant’s terminal can’t show a hidden password prompt, it will offer two choices: run the command yourself, or explicitly allow it to pass a password you provide for that one step. Passwords are never written to files, arguments, or logs.

Authenticated commands fail with a credential-store error — show / hide

The assistant’s sandbox may not reach your operating system’s credential store. Ask it to rerun the same command outside the sandbox. Only if that also returns AUTH_REQUIRED do you need to sign in again.

The assistant stops and asks which app package — show / hide

Your repository has more than one app package, and the skill won’t guess. Name the package in your request. If both app-pkg/ and root-level packages exist, give it the package root as well.

A fetch says the local file differs from the server — show / hide

Your local copy has changes the server doesn’t. Commit or compare them first. The skill asks before it uses --force, which replaces your local file.

A save is rejected because the remote DRAFT is newer — show / hide

Someone changed the DRAFT after you last fetched it. Fetch and compare, or, if you’ve reviewed that version and mean to replace it, save with --overwrite-remote-draft --expected-remote-version <version>.

Validation fails right after an expression edit — show / hide

Shell quoting can strip quotes or empty strings from inline JSON. Ask the assistant to use a file-backed patch (--inputs-patch @file) and to read the node back before running tests. Quote-free guards such as {{!!...}} avoid the problem.

“This operation can’t be performed in the configured environment” — show / hide

Your server doesn’t support a feature the request needs, such as workflow CAPABILITY nodes. The skill checks this with get-server-compatibility. If the server release and skill release don’t match, update the skill to the matching release rather than working around it.

The workflow ran, but a value is empty — show / hide

Check the expression in order: no spaces inside the braces, a real node code rather than a display name, a node that runs before this one, and a path that matches that node’s actual output.

The assistant built against the wrong user’s data — show / hide

Don’t let it assume the service account or aistudio.fa-user is the end user. Ask how the runtime user maps to the Fusion identity the data needs, and to flag it as a dependency if the contracts don’t say.

I asked it to publish a workflow and it refused — show / hide

That’s by design. Workflows are published in AI Agent Studio or through your CI/CD pipeline, never from the CLI.

Key terms

  • AI Studio CLI: The command-line tool that manages an AI Agent Studio project: scaffold, fetch, create, modify, validate, test, debug, and save.
  • aistudio skill: The package a coding assistant loads to work on AI Agent Studio projects. Contains routing rules, references, and the bundled CLI.
  • Project root: The folder with env.properties, the skill, and your artifacts. Every command runs from here.
  • env.properties: Your personal connection settings for one Fusion environment. Never committed.
  • Artifact: A workflow, app, agent, Business Object, tool, topic, deeplink, or other AI Agent Studio object, stored locally as a file.
  • App package: A source-controlled layout marked by sources/app-package.json, with sources, tests, and reports per module.
  • App-backed workflow: A workflow that serves an Agentic App panel and routes stages on $OraMessageHint.
  • Material edit: A change to what an artifact does. Triggers validation and a test sync.
  • Pre-build checkpoint: The design summary the assistant shows and waits on before creating an app’s artifacts.
  • DRAFT: The unpublished server version of an artifact. Saves update it; they never publish.
  • Debugger sidecar: Local state from the last debug run: path, node results, pins, and overrides.

Previous: Best Practices for Building with the AI Studio CLI
Back to the Learning Path for Fusion AI Agent Studio CLI