Agent tools
Create an image from a concrete brief.
Generate an image from a text prompt.
What the agent can use
Generate image
Describe the subject, composition, style and intended use. Check the API catalog for the image options the current endpoint supports.
Generate image
Generate an image from a text prompt.
Try a task
Make the result easy to verify.
Choose an agent and review its enabled tools. Capability calls can incur usage charges in addition to model and runtime costs; check the current catalog before running a batch.
Provide the input
Describe the subject, composition, style and intended use. Check the API catalog for the image options the current endpoint supports.
Use this example request
Generate a calm editorial illustration of a small team reviewing an agent’s work. Use a light background, clear focal point and no text.
Inspect the actual output
Open the actual generated image. Check composition, unwanted text, visual artifacts and whether the output fits its destination.
Fit and limits
Check the tool against the task.
A written description is not an image artifact. Provider examples may expose options that are not available through the current AgentSky endpoint.
Direct API integration
The API catalog lists the currently exposed direct endpoints, input fields, limits and pricing. The agent capability set and direct API catalog are different surfaces; not every capability has an equivalent direct endpoint. Provider documentation describes the provider API; it does not guarantee that every provider option is exposed through AgentSky.
A failed call is not a result
If the call fails, inspect the error and required input before retrying. Do not treat an agent's description of an output as evidence that the tool produced it.
Common questions
Generate image questions
Is this the same as a model?
No. A model reasons about the task; a tool performs a particular operation. The selected agent must have access to the needed capability.
Where do I find request fields and prices?
Use the live tool API catalog for current endpoints, schemas, limits and pricing. Agent execution can also incur model and runtime charges.
How do I know the result is usable?
Open the actual generated image. Check composition, unwanted text, visual artifacts and whether the output fits its destination.