Terminal Agents

Terminal AI Agents - The "Headless" Powerhouses

[!TIP] These tools allow running AI Agents directly in the terminal, bypassing browser limitations and enabling direct file system access. Source: "You've Been Using AI the Hard Way" (Transcript).

1. Claude Code (The Daily Driver)

Role: Premium Autonomous Developer.

• Key Features:
  • Agents: Spawn sub-agents (e.g., "Brutal Critic", "Researcher") to work in parallel.
  • Context: Maintains claude.md for persistent project context.
  • Power: Can authenticate via Claude Pro (no API keys needed).
  • Workflow: claude (interactive) or claude -p "prompt" (headless).
• Use Case: Complex coding tasks, deep research, and multi-agent orchestration.

2. Gemini CLI (The Free Powerhouse)

Role: Accessible & Fast Terminal AI.

• Key Features:
  • Free Tier: Very generous usage limits.
  • Context: Maintains gemini.md.
  • System Access: Can read/write files, execute scripts, and analyze projects.
• Use Case: Quick answers, file manipulation, and zero-cost automation.

3. Open Code (The Local & Open Alternative)

Role: Private & Flexible.

• Key Features:
  • Local Models: Run Llama 3.2, DeepSeek, etc., locally (via Ollama).
  • Claude Auth: Can "borrow" Claude Pro auth for cloud models.
  • TUI: Nice Terminal User Interface.
  • Config: managed via config.jsonc.
• Use Case: Privacy-focused tasks, offline work, or using un-censored local models.

4. Workflow Integration (The "Holy Trinity")

Strategy: Run them all in the same directory.

• Unified Context: Ensure claude.md, gemini.md, and agents.md (Codex) are synced or referenced.
• Parallelism:
  • Use Claude for architecture/coding.
  • Use Gemini for broad research/docs.
  • Use Open Code (or Codex) for critique/review.
• Result: 3 AI brains working on the same file system simultaneously without copy-pasting.