The research workflow lets you explore technical possibilities with one or more agents before writing a feature spec. Research is the inbound funnel — it feeds into feature creation.

When to use

• You’re evaluating multiple technical approaches (e.g., “which auth library?”)
• You need diverse perspectives before committing
• The problem space is unclear and needs exploration

Lifecycle overview

All commands below work as both CLI (aigon research-start ...) and slash commands (/aigon:research-start ... or /ars ...). We show CLI form here.

Drive mode (single agent)

# Create a research topic — describe what you want to investigate
aigon research-create api-design
# → Describe: evaluate REST vs GraphQL vs tRPC for our use case...
 
# Prioritise and assign an ID
aigon research-prioritise api-design     # Assigns ID, e.g. #05
 
# Start research — agent investigates and writes findings
aigon research-start 05
 
# Agent writes findings directly into the research document
aigon research-do 05
 
# Close when done
aigon research-close 05

Fleet mode (multi-agent)

Run multiple agents to get diverse perspectives:

aigon research-start 05 cc gg cx

This does everything in one step:

• Creates separate findings files for each agent (research-05-cc-findings.md, etc.)
• Creates a tmux session for each agent
• Opens terminal windows and launches each agent with research-do
• Each agent researches independently and writes to their own findings file

The agents start researching immediately. You can monitor progress in the dashboard or attach to any session:

tmux attach -t aigon-r05-cc              # Attach to Claude's session

If a session ends or crashes, re-open it:

aigon research-open 05                   # Re-open all agents

Synthesise

After all agents submit their findings, synthesise the results:

aigon research-synthesize 05

The synthesising agent reads all findings, presents a comparison with recommendations, and asks which features to include. Use a different model than those that conducted the research for unbiased synthesis.

Complete

aigon research-close 05

The main research file contains the synthesised recommendation, with individual findings preserved in logs/.

Research autopilot

For fully hands-off research:

aigon research-autopilot 05 cc gg

This handles setup, spawning agents, monitoring, and synthesis in one command.

Output

Research output typically includes:

• Findings per agent (what they discovered)
• Trade-off analysis
• Recommended approach
• Suggested feature specs to create next