Agent tool definitions
Tool definitions for an agent framework, generated from the live OpenAPI document, so the schema a model fills is the schema the endpoint validates. format=openai (default) returns { type, function } objects you can spread straight into a chat completion; format=anthropic returns { name, description, input_schema }. The objects carry nothing of ours: how to call each tool lives in a separate executors map keyed by tool name, because an unexpected key inside a vendor tool object is a 400 at their end. select chooses what to hand over: the curated starter set by default, all for the catalogue, group:Vedic for one area, paths:/chart,/synastry for an explicit list; q filters within that, limit caps it (default 32, max 128, which is the OpenAI ceiling). Nothing truncates silently: totalMatched, totalAvailable and a note say what was left out. No auth is embedded in a tool, so send X-Api-Key on the call itself.
Query parameters
Which vendor contract the tool objects follow. openai returns { type, function }, anthropic returns { name, description, input_schema }.
What to hand over: starter (the curated set, the default), all, group:<tag> such as group:Vedic, or paths:/chart,/synastry for an explicit list. An unknown path is named in notes rather than dropped.
Free-text filter applied within the selection, matched against path, summary, description and group.
How many tools to return. The ceiling is the OpenAI limit of 128 functions per request; models degrade well before it. Anything left out is counted in totalMatched.
Response
Successful calculation
Example response
{
"ok": true
}