# Tools

Use these guides to prepare a task and check its output. For direct API calls, the live catalog owns endpoint inputs, limits and pricing.

### Web search
Use Exa search to find public URLs for a question, and Exa contents to retrieve the text of a known page. A useful research task combines both: discover relevant sources, read the strongest ones, and produce a comparison whose claims link back to those sources.

### Browser use
Use browser-backed retrieval for a JavaScript-heavy directory, product listing or application page that a plain fetch cannot read. Give the agent the starting URL and the fields you need, such as product name, visible availability and source link. These capabilities are classified as read operations in the agent catalog.

### Image generation
AgentSky uses OpenAI gpt-image-2 to generate one image at high quality. Specify the subject, composition and exact wording, then choose a square, portrait or landscape size.

### SEO data
Use the DataForSEO Google Organic Live Advanced endpoint to study the pages competing for a search query. The current AgentSky adapter fixes the location to the United States and derives the language from locale. This makes the tool useful for US search-result research, but not for choosing an arbitrary country.

### Background removal
Remove the surrounding background while keeping the source subject. This fits catalog images, presentation assets and profile cutouts where you already have the right photograph. It does not invent a new setting or replace objects; those jobs require image editing or generation.

### Text to video
Describe one coherent shot: the setting, subject action and camera movement. The current adapter defaults to ByteDance Seedance 2.0, submits an asynchronous generation job and returns an MP4 when the job succeeds. The configured default can be overridden by the deployment.

### Image to video
Start with a publicly reachable image URL and describe the motion you want. The current adapter defaults to Alibaba Wan 2.6 I2V Flash and requests a 1280×720 video. The historical mm.i2v identifier does not indicate a MiniMax model.

### Transcription
The current adapter uses Fish Audio transcribe-1. Supply audio as an accessible URL or a local file through the CLI, and use automatic language detection or a language code. The returned transcript can feed a separate summarization task, such as extracting decisions and action items from an interview.

## Tools, models and connectors
### How are tools different from models and connectors?
Models reason and generate responses. Tools perform operations such as search or image generation. Connectors authorize access to an application account. A workflow can require all three.

### Where is the complete tool listing?
Open the live tool catalog for the currently served endpoints. These task guides are curated explanations, not a replacement for that listing.