intellectronica

Eliteagent — AI skill for Claude Code

AI community

Terminal AI coding agent powered by pydantic-ai, rich, and prompt-toolkit.

How to install Eliteagent

This entry records only its repository, not the path inside it, so there is no exact command to give. Open intellectronica/eliteagent and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Eliteagent does

Terminal AI coding agent powered by pydantic-ai, rich, and prompt-toolkit.

Alternatives in AI

  • Nodeterm — Node-based terminal manager for AI coding agents — tmux-backed terminals and parallel agent sessions as dragga 1.3k ★
  • Session Kit — Close the terminal windows, not the AI coding sessions 649 ★
  • Dinotty — Multi-device terminal server for AI coding agents 643 ★

README

EliteAgent

Terminal AI coding agent powered by pydantic-ai, rich, and prompt-toolkit.

Features

  • Multiple models: OpenAI GPT-5, Anthropic Claude 4.5 Sonnet, OpenRouter Grok Code Fast
  • Rich UI: boxed messages, color-coding, spinner
  • Tool: shell for executing commands (stdout+stderr)
  • Modes: Approval vs YOLO
  • Ctrl+C handling & session history

Requirements

  • Python 3.10+
  • API keys set in .env (see example.env)

Quick start

# Install uv (optional but recommended)
# macOS: brew install uv

# Create env and run
uv sync
# Run via module to avoid PATH script resolution issues
uv run -m eliteagent.cli

If not using `uv`:

python -m venv .venv
source .venv/bin/activate
pip install -e .
# If `eliteagent` is not on PATH in your shell, run via module:
python -m eliteagent.cli

Environment

Create a `.env` in the project root from `example.env` and set:

OPENAI_API_KEY=...
ANTHROPIC_API_KEY=...
OPENROUTER_API_KEY=...
# Optional OpenRouter endpoint override
# OPENROUTER_BASE_URL=https://openrouter.ai/api/v1

Usage

  • Type your request at the prompt.
  • Slash commands:
    • /new — new session (clears history)
    • /model — switch models
    • /approval — toggle Approval vs YOLO for tool execution

Notes

  • The shell tool executes via /bin/bash -lc to support pipes and redirection.
  • All model text messages are displayed; tool calls and outputs are boxed and color-coded.