Research Agent
The autonomous research agent browses, searches, and extracts information based on a natural language prompt. Use it for complex multi-site research that would be tedious to do manually.
Tools
| Tool |
Description |
agent |
Start an autonomous web research agent |
agent_status |
Check agent results |
Start Research Agent
Tool: agent
Input: {
"prompt": "Find the pricing plans for the top 5 project management tools and compare their features",
"urls": ["https://asana.com/pricing", "https://monday.com/pricing"]
}
Returns: { "agent_id": "agent_xxx", "status": "running" }
Parameters
| Parameter |
Type |
Description |
prompt |
string |
Research task description (required, max 10,000 chars) |
urls |
string[] |
Starting URLs to guide the agent (optional) |
schema |
object |
JSON schema for structured output (optional) |
Check Agent Results
Tool: agent_status
Input: { "id": "agent_xxx" }
Returns: {
"status": "completed" | "running" | "failed",
"result": { ... }
}
Workflows
Research with Structured Output
1. agent({
prompt: "Research the top 5 CRM tools, find their pricing and key features",
schema: {
"type": "object",
"properties": {
"tools": {
"type": "array",
"items": {
"type": "object",
"properties": {
"name": { "type": "string" },
"pricing": { "type": "string" },
"features": { "type": "array", "items": { "type": "string" } }
}
}
}
}
}
}) → Start agent
2. agent_status(agent_id) → Poll until complete
3. Process structured results
Research Starting from Known URLs
1. agent({
prompt: "Compare the API documentation quality of these services",
urls: ["https://docs.stripe.com", "https://docs.paypal.com"]
}) → Start with specific URLs
2. agent_status(agent_id) → Get comparison results
When to Use Agent vs Manual Tools
| Use Case |
Approach |
Why |
| Simple page scrape |
scrape_url |
Faster, more predictable |
| Known data extraction |
extract_data |
Direct, schema-based |
| Complex multi-site research |
agent |
Autonomous navigation |
| Open-ended exploration |
agent |
Agent decides what to search/scrape |
| Comparison across many sources |
agent |
Agent handles discovery |
Best Practices
- Be specific in prompts — clear task description gives better results
- Provide starting URLs — guides the agent to relevant sources
- Use schemas — structured output is more reliable than free-form
- Keep prompts under 10k chars — respect the limit
- Poll for results — agents take time for complex research
Error Handling
- Agent fails → simplify prompt or provide more specific URLs
- Incomplete results → break research into smaller, focused prompts
- Schema mismatch → ensure schema properties match what can be found on the web
- Timeout → reduce scope of research task
agents-store