B
Offline
Step 1 of 10

Configure Agent Subscriptions

2of 10~10 min
  • API key management across providers
  • Rate limits and cost considerations
  • How to verify agent access
What you'll do
  • Verify Claude Code / Claude Max access
  • Verify Codex / GPT Pro access
  • Verify Gemini CLI / Gemini Ultra access

Multi-agent research requires active subscriptions to three AI providers. Each agent brings unique strengths to the research process.

Cost Consideration: Running all three agents for a full session (1-2 hours) typically costs $5-15 in API usage, depending on context length and iteration count. Monitor your usage dashboards.

C

Claude

Opus 4.5 • Claude Max

Claude Code is Anthropic's coding agent CLI. It requires a Claude Max subscription for access to Opus 4.5 (the most capable model).

Install & Verify

Terminal
# Install Claude Code (if not already installed)
# See: https://github.com/anthropics/claude-code
# Verify installation
claude --version
# Authenticate (opens browser)
claude auth
# Test with a simple prompt
claude "What is 2 + 2?"
Best for: Test design, careful reasoning, nuanced analysis
C

Codex

GPT 5.2 • GPT Pro

Codex CLI is OpenAI's coding agent. GPT Pro subscription unlocks extended context and higher rate limits needed for research sessions.

Install & Verify

Terminal
# Install Codex CLI (if not already installed)
# See: https://github.com/openai/codex-cli
# Verify installation
codex --version
# Set your API key
export OPENAI_API_KEY="your-key-here"
# Test with a simple prompt
codex "What is 2 + 2?"
Best for: Creative hypotheses, broad knowledge, cross-domain connections
G

Gemini

Gemini 3 • Gemini Ultra

Gemini CLI is Google's coding agent. Gemini Ultra provides access to the most powerful Gemini model with extended context.

Install & Verify

Terminal
# Install Gemini CLI (if not already installed)
# See: Google AI Studio for API access
# Verify installation
gemini --version
# Set your API key
export GOOGLE_AI_API_KEY="your-key-here"
# Test with a simple prompt
gemini "What is 2 + 2?"
Best for: Adversarial critique, edge-case finding, skeptical analysis

✓Verify All Agents

Run this script to verify all three agents are accessible:

check-agents.sh
#!/bin/bash
echo "=== Multi-Agent Prereq Check ==="
# Check Claude Code
claude --version && echo "✓ Claude Code" || echo "✗ Claude Code missing"
# Check Codex
codex --version && echo "✓ Codex" || echo "✗ Codex missing"
# Check Gemini CLI
gemini --version && echo "✓ Gemini CLI" || echo "✗ Gemini CLI missing"
echo "=== End Check ==="
Pro Tip
If budget is a concern, you can start with just one or two agents. Claude + Codex is a strong pair for hypothesis generation and test design. Add Gemini for the full adversarial critique capability.

All Agents Ready? If all three CLI tools respond to --version, you're ready to define your research roster in the next step.