Search and Retrieval APIs
Use Tavily with LlamaIndex
Tavily connects to LlamaIndex through a official adapter package path (llama-index-tools-tavily-research (TavilyToolSpec)). This page records what that connection looks like and links the evidence behind it.
Pairing facts
Why this pairing
A web access API built for AI agents: search, extract, map, crawl, and research endpoints behind one key.
Best for: Agents that need current web facts with cited results Research workflows that combine search with page extraction Teams that want one vendor for search and fetching
Not ideal for: Fully offline or self-hosted stacks High-volume crawling without a credit budget
Configuration
pip install llama-index-tools-tavily-research
from llama_index.tools.tavily_research import TavilyToolSpec
tavily_tool = TavilyToolSpec(api_key=os.getenv("TAVILY_API_KEY"))Source: https://developers.llamaindex.ai/python/framework/understanding/agent/streaming/. Examples use placeholders only.
Evidence
The LlamaIndex developer docs streaming page installs llama-index-tools-tavily-research, imports TavilyToolSpec from llama_index.tools.tavily_research and uses TAVILY_API_KEY.
https://developers.llamaindex.ai/python/framework/understanding/agent/streaming/
Compatibility here means the cited page documents the pairing. It does not mean AgentsUse benchmarked the combination or that every version works unchanged. Pin versions, run the tool on a small job first, and review the first outputs before widening access.
Watch for
- Metered by credits, so unattended crawl loops need spending controls.
- A managed service: no self-hosted fallback when Tavily is down or a source blocks it.