Research

The research feature allows the AI agent to conduct in-depth research from multiple sources and generate comprehensive reports.

Overview

Research jobs gather information and synthesize findings:

Creating Research Jobs

From Chat

Ask the agent to research:

You: Research best practices for API rate limiting

Agent: I'll research API rate limiting best practices.
       Sources: Documentation, Web
       Estimated time: 5-10 minutes

       [Starting research...]

Research Configuration

When creating research, specify:

Option	Description	Default
Topic	What to research	Required
Sources	Where to look	Docs + Web
Time Budget	Max research time	60 minutes
Output Mode	Report format	Longform
Target Space	Where to save report	Current

Research Sources

Documentation

Search your workspace documentation:

• All accessible pages
• Linked attachments
• Page history
• Comments and discussions

Web

Search the public internet:

• Technical documentation
• Blog posts and articles
• Stack Overflow
• Official guides

Repositories

Analyze GitHub repositories:

Research Config {
  sources: ['repository'],
  repository: {
    url: 'https://github.com/org/repo',
    paths: ['src/', 'docs/'],  // optional filter
    branch: 'main',
  }
}

Repository research can analyze:

• Code structure
• README and documentation
• Implementation patterns
• Dependencies

Output Modes

Longform

Comprehensive report with:

• Executive summary
• Detailed findings
• Code examples
• References
• Recommendations

Best for: Deep dives, documentation, decision support

Brief

Concise summary with:

• Key points
• Quick recommendations
• Essential links

Best for: Quick answers, status updates, overviews

Research Workflow

1. Request

Clearly state what you need:

Good: "Research authentication patterns for Node.js APIs,
       focusing on JWT vs session-based auth"

Too vague: "Research authentication"

2. Source Selection

Choose appropriate sources:

Need	Recommended Sources
Internal knowledge	Docs only
Industry practices	Web
Code patterns	Repository + Web
Comprehensive	All sources

3. Time Budget

Set realistic time budgets:

Research Type	Suggested Time
Quick lookup	5 minutes
Topic overview	15-30 minutes
Deep dive	60+ minutes
Comprehensive	120+ minutes

4. Review Results

Research generates a report page:

# Research: API Rate Limiting Best Practices

## Summary
[Executive summary of findings]

## Key Findings

### 1. Token Bucket Algorithm
[Details and examples]

### 2. Sliding Window
[Details and examples]

## Recommendations
[Actionable recommendations]

## Sources
[List of sources consulted]

---
Generated: 2024-01-15
Time spent: 45 minutes
Sources: 12 documents, 8 web pages

Managing Research Jobs

View Active Research

See ongoing research:

1. Open Agent Chat
2. Click Active Jobs
3. View progress and status

Research History

Access past research:

1. Navigate to the target space
2. Look for research report pages
3. Or search "Research:" in your workspace

Cancel Research

Stop ongoing research:

You: Cancel the current research

Agent: Research job cancelled. Partial results saved.

Advanced Features

Multi-Source Synthesis

The agent combines information from multiple sources:

Follow-up Research

Dive deeper into findings:

You: Tell me more about the token bucket algorithm
     mentioned in the research

Agent: [Provides detailed explanation or starts
       focused follow-up research]

Research Templates

Create templates for common research:

Research Template: "Competitive Analysis" {
  sources: ['web'],
  outputMode: 'longform',
  timeBudget: 90,
  structure: [
    'Market Overview',
    'Key Players',
    'Feature Comparison',
    'Recommendations',
  ]
}

API Access

Create Research Job

const job = await client.research.create({
  workspaceId: 'ws_123',
  topic: 'API rate limiting best practices',
  sources: ['docs', 'web'],
  outputMode: 'longform',
  timeBudget: 60,
  targetSpaceId: 'space_456',
});

Check Status

const status = await client.research.status({
  workspaceId: 'ws_123',
  jobId: job.id,
});

// status: 'pending' | 'running' | 'completed' | 'failed'

Get Results

const report = await client.research.getReport({
  workspaceId: 'ws_123',
  jobId: job.id,
});

Best Practices

1. Be specific - Clear topics yield better results
2. Choose sources wisely - Match sources to the question
3. Set time budgets - Balance depth with urgency
4. Review and refine - Use findings to ask follow-ups
5. Save valuable research - Link reports to relevant projects

Limitations

• Web research respects robots.txt
• Repository analysis requires public access or auth
• Time budgets are approximate
• Very large repositories may be partially analyzed

Related

• AI Agent - Agent capabilities
• Memory System - Research stored in memory
• MCP Research Tools - API access