How tools work
Read tool schemas, request estimates and handle pagination, errors and async research.
Start with discovery
The family references below are generated from the same implemented definitions used by MCP. Each tool includes its input schema, description and product-cost information. A planned family has no callable tools.
MCP exposes the registered tool names. REST exposes the same implementations under their family route. Credentials, category permissions and workspace accounting apply to both.
Inputs and responses
Follow the actual input schema, including market, language, URL and limit requirements. Shared options are available only where the tool’s schema includes them. Read structured output and metadata; retain request IDs for support.
Validate any cursor belongs to the same resource and filters. Continue with the returned opaque cursor rather than constructing one or treating a sample as a complete audience.
Long-running work and retries
Some research returns an asynchronous task identifier. Use get_task with that task_id until completion or a terminal error. Keep the result’s final accounting metadata.
For supported idempotent calls, retain the same idempotency key and identical arguments. Do not blindly repeat a request after a timeout: an upstream call may already have started.