APIs your agents can find on their own.
Point a client at it and your agent searches this marketplace before it builds anything. Hit, and it runs it. Miss, and it describes what it needs — then gets a typed function back.
Builds started here are published automatically with their source code. Other users can run, download and fork them. Check your build instructions for private information. Unpublishing cannot retract downloaded or forked copies.
Returns per-model Intelligence and Coding Index scores, task cost breakdowns, timing, latency, TTFT, and token counts from https://artificialanalysis.ai.
Returns per-model Intelligence and Coding Index scores, task cost breakdowns, timing, latency, TTFT, and token counts from https://artificialanalysis.ai. It runs on Notte's browser infrastructure and returns structured JSON, so you can call it like any other API — pass models and get the result back without maintaining a scraper.
Hand the whole endpoint to Claude Code or any MCP client instead of writing the call yourself.
curl --location 'https://us-prod.notte.cc/functions/0cfd4528-4af5-4b55-9b23-07b0c78c0c13/runs/start' \
--header 'x-notte-api-key: YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer YOUR_API_KEY' \
--data '{
"function_id": "0cfd4528-4af5-4b55-9b23-07b0c78c0c13",
"variables": {
"models": ""
}
}'{
"count": 0,
"records": [
{
"id": "<string>",
"name": "<string>",
"slug": "<string>",
"creator": {
"id": "<string>",
"logo": "<string>",
"name": "<string>",
"slug": "<string>",
"color": "<string>",
"country": "<string>"
},
"short_name": "<string>",
"ttft_seconds": 0,
"cost_per_task_usd": 0,
"coding_index_score": 0,
"cost_breakdown_usd": {},
"time_per_task_seconds": 0,
"token_counts_per_task": {
"input": 0,
"answer": 0,
"output": 0,
"reasoning": 0
},
"ttft_breakdown_seconds": {
"input": 0,
"total": 0,
"reasoning": 0
},
"coding_cost_per_task_usd": "<any>",
"intelligence_index_score": 0,
"cost_per_task_by_evaluation": [
{
"slug": "<string>",
"weightedCostPerTask": 0
}
],
"coding_time_per_task_seconds": "<any>",
"intelligence_index_is_estimated": true,
"end_to_end_response_time_seconds": 0,
"median_time_to_first_chunk_seconds": 0,
"end_to_end_response_time_breakdown_seconds": {
"input": 0,
"total": 0,
"answer": 0,
"reasoning": 0
}
}
],
"requested_models": [
"<string>"
]
}