search_apis
Find APIs that match your natural language task description, with optional cost and latency constraints.
Instructions
DEPRECATED in v0.1.7. Use 'resolve' instead. This alias will be removed in v1.0.0. [Original behavior preserved in this version: Find the best API for a task using natural language.]
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| budget | No | DEPRECATED in v0.1.9. Prefer cost_max_per_call_usdc. Maximum USDC cost per call (e.g. 0.01). Invalid values silently ignored. When both budget and cost_max_per_call_usdc are set, the stricter (smaller) value applies. | |
| intent | Yes | What you want to do in natural language (e.g. 'summarize news articles', 'get token price') | |
| latency | No | Required response latency tier | |
| min_similarity | No | Similarity threshold for vector search. Default 0.5. Out-of-range returns 400. | |
| latency_p95_max_ms | No | v0.1.9 (D-084): Maximum measured p95 latency in milliseconds. Invalid values (≤0, NaN) return 400. APIs with no measured p95_latency_ms are excluded from results when this filter is set. | |
| cost_max_per_call_usdc | No | v0.1.9 (D-084): Maximum cost per call in USDC. Invalid values (negative, NaN) return 400. When both budget and this are set, the stricter (smaller) value applies. |