tavily-research

Tavily Research Skill

Conduct comprehensive, citation-backed research on any topic. Tavily performs multiple iterative web searches, synthesizes sources, and returns a structured report — no manual source gathering required.

Authentication

Set TAVILY_API_KEY in ~/.openclaw/.env.local:

TAVILY_API_KEY=tvly-your-api-key-here

When to Use This Skill

• Multi-source comparisons ("LangGraph vs CrewAI vs AutoGen")
• Market analysis or competitive landscape reports
• Current events synthesis with citations
• Technical deep-dives that need multiple authoritative sources
• Any question where a single search result is insufficient

Prefer tavily-search for quick single-topic lookups where you just need a list of results.

Note: Research takes 30–120 seconds. Set user expectations accordingly before starting.

API Overview

The Research API is asynchronous: you submit a task, then poll for results.

Step 1 — Submit research task:

POST https://api.tavily.com/research

Step 2 — Poll for status/results:

GET https://api.tavily.com/research/{task_id}

Step 1: Submit Task

curl -s -X POST https://api.tavily.com/research \
  -H "Authorization: Bearer $TAVILY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "Compare LangGraph vs CrewAI for multi-agent systems in 2025",
    "model": "pro"
  }'

Response:

{
  "task_id": "res_abc123xyz",
  "status": "pending"
}

Request Fields

Field	Type	Default	Description
input	string	Required	Research topic or question
model	string	"mini"	mini / pro / auto

Model Selection

Model	Best For	Typical Time
mini	Single topic, targeted question	~30s
pro	Multi-angle analysis, comparisons, reports	60–120s
auto	Let Tavily decide based on complexity	Varies

Rule of thumb: "What does X do?" → mini. "X vs Y vs Z" or "best way to..." → pro.

Step 2: Poll for Results

curl -s -X GET https://api.tavily.com/research/res_abc123xyz \
  -H "Authorization: Bearer $TAVILY_API_KEY"

Response when complete:

{
  "task_id": "res_abc123xyz",
  "status": "completed",
  "report": "## LangGraph vs CrewAI\n\n...",
  "sources": [
    {
      "title": "LangGraph Documentation",
      "url": "https://langchain-ai.github.io/langgraph/",
      "content": "..."
    }
  ]
}

Poll every 5–10 seconds until status is "completed" or "failed".

Examples

Quick targeted research (mini)

# Submit
curl -s -X POST https://api.tavily.com/research \
  -H "Authorization: Bearer $TAVILY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "What is retrieval augmented generation?", "model": "mini"}'

# Poll with task_id from response
curl -s https://api.tavily.com/research/{task_id} \
  -H "Authorization: Bearer $TAVILY_API_KEY"

Comprehensive market research (pro)

curl -s -X POST https://api.tavily.com/research \
  -H "Authorization: Bearer $TAVILY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "Fintech startup landscape 2025 key players funding trends", "model": "pro"}'

Tips

• Always tell the user research is in progress — it takes time
• Use mini for "explain X" style questions, pro for analysis/comparison
• The returned report is already synthesized markdown with inline citations
• Check sources array to verify and share references with the user
• If status is "failed", retry once with a simpler or more specific input