A saved tool contains a reusable workflow. A deployment is the configured placement you run from the Tools page or another surface.
Reusable Tool
The reusable tool includes:
- Its name and description.
- The original intent used to create it.
- Tags and organization.
- An ordered list of steps.
- Default settings and instructions for new deployments.
Every current tool is a fixed linear workflow. Steps run in order; the tool does not branch based on a condition.
Deployment
A deployment controls:
- Where the tool appears.
- Which project owns the placement.
- Its display name.
- Documents and folders used as input.
- Its settings and instructions.
- Its AI model route.
- Its schedules and folder watches.
The same reusable tool can have several independent deployments.
Input and Step Results
A run begins with the input supplied by the current surface. Each step receives the original input and the preceding result. The last completed step provides the final result.
A folder or multi-file input can be combined into one run or processed one file at a time when Process each input file separately is enabled.
Autonomous and Guided Modes
An AI-backed tool can run in autonomous or guided mode.
Guided mode requests approval after AI steps. Desktop and mobile runs can present the request. Unattended runs, including triggers, approve it automatically.
Tool Settings
The fields shown in Settings depend on the tool. They can include identity, instructions, execution behavior, connected-service options, output preferences, and skill-specific values.
A Python Script Tool can show its interpreter, script, arguments, working folder, input mode, timeout, and trust acknowledgment.
Settings have two scopes:
- Settings opened on the reusable tool change the defaults copied into future deployments.
- Tool settings opened from a selected deployment changes only that deployment.
For example, two deployments of the same Python tool can use different scripts, interpreters, arguments, and timeouts. Changing one does not change the other.
The Settings dialog does not edit the tool's step list.
AI Model
AI controls appear only when the workflow uses AI.
On a selected deployment, Change model shows the provider and model used by that deployment.
- Select Use Default Profile to follow the application's active AI profile.
- Clear Use Default Profile to save a provider and model override for this deployment.
- When OpenRouter is selected, you can also choose the model provider.
A deployment override does not change the global AI profile or another deployment.
Details
The inspector's collapsed Details section shows placement and technical information. These values are useful for troubleshooting but are not required for normal use.