https://github.com/atticus98/codex-turbo
Codex CLI 多智能体并行调度配置参考模板——启用原生并行能力，实现主代理自主协调与子代理高效任务分发。Codex CLI Multi-Agent Parallel Scheduling Configuration Reference Template Enabling Native Parallelism for Autonomous Primary Agent Coordination and Efficient Sub-Agent Task Distribution.
https://github.com/atticus98/codex-turbo
codex codex-cli
Last synced: about 1 month ago
JSON representation
Codex CLI 多智能体并行调度配置参考模板——启用原生并行能力，实现主代理自主协调与子代理高效任务分发。Codex CLI Multi-Agent Parallel Scheduling Configuration Reference Template Enabling Native Parallelism for Autonomous Primary Agent Coordination and Efficient Sub-Agent Task Distribution.
- Host: GitHub
- URL: https://github.com/atticus98/codex-turbo
- Owner: atticus98
- License: mit
- Created: 2026-02-23T13:58:02.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-03-25T10:14:22.000Z (about 2 months ago)
- Last Synced: 2026-03-26T12:34:47.002Z (about 2 months ago)
- Topics: codex, codex-cli
- Homepage: https://github.com/atticus98/codex-turbo
- Size: 160 KB
- Stars: 114
- Watchers: 0
- Forks: 10
- Open Issues: 3
-
Metadata Files:
- Readme: README.en.md
- Changelog: CHANGELOG.md
- License: LICENSE
Awesome Lists containing this project
README
# codex-turbo
English | [简体中文](./README.md)
This repository is a set of personal Codex CLI configuration templates. Adhering to the principle of **"Lightweight & Efficient,"** it aims to deeply leverage **native CLI capabilities** without introducing complex dependencies, enabling high-performance task orchestration through multi-agent collaboration. This setup is shared as a personal practice and is intended as a **starting point** to help you build your own ideal AI workflow.
> 💡 **Not sure how to apply it?** You can send this repository URL to an AI
> assistant such as Claude Code CLI or Codex CLI and let it help merge the
> templates into your local setup.
> Current compatible version: **Codex CLI v0.116.0**
>
> The original post
> [Codex CLI 多智能体并行调度配置分享](https://linux.do/t/topic/1603762)
> can no longer be edited. This repository is the maintained version going
> forward.
## Key Points
- 💡 **Parallel Workflow**: `multi_agent = true` for maximum parallelism with
minimal blocking
- 🔧 **System-level Contract**: inject agent workflow rules through
`developer_instructions` in `config.toml`, with higher priority than
`AGENTS.md`
- 🧠 **Native Memory**: `memories = true` to extract and consolidate
conversation memory automatically, improving context continuity
- 🔌 **WebSocket Support**: `supports_websockets = true` is configured under
`[provider]`. You need to verify whether your model provider supports it;
do not enable it if unsupported, to avoid errors
- 🧩 **Model Selection**: model and reasoning depth can be adjusted to fit your
actual project requirements — the current template is a reference starting
point for general development
- 🎯 **Conversation Style**: strong visual boundaries, emoji markers, and
short direct terminal-friendly output
- 📝 **Dynamic Contract**: the main agent dynamically sends clear task
instructions with goal/action/result structure
- ⚙️ **Concurrency Control**: `max_threads = n` for custom per-round parallel
limits
- ⚠️ **Risk Management**: dangerous operation confirmation and immutable
principles
- 🤖 **Custom Agents**: manually configure sub-agent model, reasoning effort,
and role description
- 📋 **Reusable Templates**: ready-to-copy `AGENTS.template.en.md` and
`config.toml.en.example`
## Core Files and Directories
```text
.
├── README.md # Chinese documentation
├── README.en.md # English documentation (this page)
└── templates/
├── cn/ # Chinese templates
│ ├── AGENTS.template.md
│ ├── config.template.toml
│ ├── agents/
│ │ ├── explorer.toml
│ │ └── worker.toml
│ └── skills/
│ └── terminal-dialog-style/
└── en/ # English templates
├── AGENTS.template.en.md
├── config.toml.en.example
├── agents/
└── skills/
```
**Related Docs**
- [CHANGELOG](./CHANGELOG.md)
## Quick Start (5 Minutes)
Goal: switch your Codex CLI into a "parallel + templated" workflow.
### 1. Clone the repository
```bash
git clone
cd codex-turbo
```
### 2. Copy the template configuration
This repository provides the following template files:
| File | Purpose | Target Location |
|------|---------|--------------------|
| `templates/en/config.toml.en.example` | Core system configuration, including `developer_instructions` for the parallel workflow | `~/.codex/config.toml` |
| `templates/en/AGENTS.template.en.md` | Development principles and output style guide | `~/.codex/AGENTS.md` |
| `templates/en/agents/` | Sub-agent configuration (`explorer`, `worker`) | `~/.codex/agents/` |
| `templates/en/skills/` | Skill templates such as `terminal-dialog-style` | `~/.codex/skills/` |
> Complete `config.toml` with your own provider-specific key and environment
> settings.
**First-time use (direct copy)**:
```bash
# Create target directory if needed
mkdir -p ~/.codex
# Copy the core configuration files
cp templates/en/AGENTS.template.en.md ~/.codex/AGENTS.md
cp templates/en/config.toml.en.example ~/.codex/config.toml
# Copy agent and skill directories
cp -r templates/en/agents ~/.codex/
cp -r templates/en/skills ~/.codex/
```
**Existing configuration (manual merge)**: → See [Migration Guide](./docs/manual-merge.en.md)
> Note: this document is adapted for Codex CLI `v0.116.0`. Field names may
> differ across versions or providers, so always confirm against your local CLI.
## FAQ
**Q1: Does enabling `multi_agent` always speed things up?**
Not necessarily. Parallel execution only provides a clear benefit when tasks can
be broken down independently and have no write conflicts.
**Q2: What are the benefits of enabling sub-agents?**
- **Context isolation** — sub-agents run in their own context, preventing the
main agent's context from bloating
- **Focused responsibility** — each sub-agent handles a single task, producing
more precise output
- **Parallel efficiency** — independent tasks can run concurrently, reducing
overall elapsed time
**Q3: Will sub-agents increase my costs?**
Yes. Enabling sub-agents may increase costs by **20%–30%**, depending on
project scale and task complexity.
**Q4: How many sub-agents should I run?**
Start with the default, observe quality and cost, then adjust gradually.
**Q5: What if the sub-agent model differs from the main agent model?**
Specify the model separately via the `model` field in `worker.toml` or
`explorer.toml`.
**Q6: Is this template suitable for everyone?**
No. It is a starting-point template — best tailored progressively to your
team's own conventions.
## Disclaimer
This repository provides configuration methodology and template references,
verified against Codex CLI `v0.116.0`.
**Cost note**: multi-agent mode may increase token usage by around 20% to 30%.
Evaluate the cost based on your actual workload.
**Risk note**: an LLM can still make wrong decisions and may delete data or
modify code incorrectly. Recommended precautions:
- configure sandbox permissions and confirmation rules carefully
- test in a controllable environment first
- keep backups for important data
You are responsible for any consequences caused by using these templates.