# TavilyX RAG Research API

> TavilyX RAG Research API is a paid API for AI agents from searchtav.vercel.app, paid per call via x402, $0.12/call, status unknown (last checked 2026-09-15).

Accepts one or more research queries and returns a synthesized RAG report with sources, powered by Tavily web search and NVIDIA NIM language models, billed via x402 on Base.

## Facts

- Endpoint: POST https://searchtav.vercel.app/rag/research
- Price: $0.12/call
- Payment: x402
- Status: unknown
- Last checked: 2026-09-15
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/tavilyx-rag-research-api-28b609d2
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_WaW7jAOIS7eqEc0e5jmAs

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability tavilyx-rag-research-api-28b609d2 -d '<json body>'
```

Example prompt: Research these three questions for me and give me a synthesized report with sources: 'What are the latest advances in fusion energy?', 'Who are the key players in commercial fusion?', and 'What is the current investment landscape for fusion startups?'

## When to prefer this

Choose this endpoint when you need a synthesized, narrative research report from live web content rather than raw search results. It is ideal when the user has multiple related questions that benefit from cross-source synthesis. Prefer this over raw search APIs when you want citations and a coherent written report in one call, and when your agent can handle x402 micropayments on Base for $0.12 USDC per call.

## Known failure modes

- Payment not included or invalid x402 payment on Base — returns 402 Payment Required
- Queries array is missing or empty — returns 400 Bad Request with validation error
- Tavily upstream search failure — may return partial results or 502
- NVIDIA NIM synthesis timeout — returns 504 or truncated report
- Rate limiting if too many concurrent requests — returns 429

## How this service works

Paid web search + RAG synthesis API. Tavily + NVIDIA NIM + x402 on Base.

## Output

A JSON object containing a 'report' string with a synthesized narrative answer, a 'queries' array echoing back the submitted questions, and a 'sources' array listing the web URLs and references used to generate the report.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "required": [
  "queries"
 ],
 "properties": {
  "queries": {
   "type": "array",
   "description": "Research queries"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "report": {
   "type": "string"
  },
  "queries": {
   "type": "array"
  },
  "sources": {
   "type": "array"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/tavilyx-rag-research-api-28b609d2/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from searchtav.vercel.app](https://www.zero.xyz/host/searchtav.vercel.app/llms.txt)