Iterative Multi-Engine AI Orchestration
Combine Claude, Gemini, and Codex in a quality-scored iteration loop. File-based event bus. Multi-agent orchestration. Zero external dependencies.
Your task enters a feedback loop. An executor builds, a reviewer scores, and the cycle repeats until quality passes your threshold.
Claude, Gemini, Codex support. Unified Engine interface. Swap providers with a flag.
Quality-scored loop. Auto-approve at threshold. Configurable max iterations and scoring.
File-based JSONL. Crash-safe. Git-friendly. Zero dependencies. Survives restarts.
Background daemon. IPC orchestration. Activity-aware routing between agents.
Executable markdown. Auto-injected prompts. Extensible skill library.
Terminal.app, iTerm2, tmux, PTY. Pluggable backends. Works everywhere.
# Install $ npm install -g @lawrence369/iter-loop # Interactive mode $ iter-loop # Direct execution $ iter-loop "Build a REST API with Express" --executor claude --reviewer gemini # Agent wrappers $ lclaude # Claude in loop mode $ lgemini # Gemini in loop mode # Management $ iter-loop daemon start $ iter-loop bus status $ iter-loop chat
┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ CLI UI │ │ Core Loop │ │ Engines │ │ @clack │────│ executor │────│ Claude │ │ blessed │ │ reviewer │ │ Gemini │ │ dashboard │ │ scoring │ │ Codex │ └─────────────┘ └──────┬──────┘ └─────────────┘ │ ┌──────┴──────┐ │ Event Bus │ │ JSONL │ │ queues │ └──────┬──────┘ │ ┌────────────┼────────────┐ │ │ │ ┌─────────┐ ┌───────────┐ ┌───────────┐ │ Agents │ │Orchestrator│ │ Terminal │ │ PTY │ │ Daemon │ │ Adapters │ │ wrapper │ │ IPC │ │ tmux/iTerm│ └─────────┘ └───────────┘ └───────────┘
| Feature | Individual CLIs | iter-loop |
|---|---|---|
| Multi-engine | Manual switching | Unified interface |
| Quality control | None | Auto-scored iteration |
| Agent coordination | Copy-paste | Event bus + IPC |
| Context sharing | Manual | Shared plan file |
| Background work | Not possible | Daemon + orchestrator |