🎱 ModelPool (Free) — Free AI Model Manager for OpenClaw
One command. Unlimited free AI. No more rate limits.
Prepare your OpenRouter API key(s) before install (free at openrouter.ai), then:
CODEBLOCK0
ModelPool will auto-discover the best free models, configure multi-key rotation, and build a smart fallback chain. You'll never hit a rate limit again.
Commands
| Command | Description |
|---|
| INLINECODE0 | The one command you need — interactive setup, enter keys, everything auto-configured |
| INLINECODE1 |
Auto-reconfigure with existing keys (non-interactive) |
|
modelpool list | Browse all free models ranked by quality score |
|
modelpool switch <model> | Manually switch primary model |
|
modelpool status | Show current config, keys, and fallback chain |
|
modelpool repair |
One-click fix — diagnoses and repairs 7 common issues automatically |
|
modelpool keys add <key> | Add another OpenRouter key (more keys = more quota) |
|
modelpool keys list | Show all configured keys |
|
modelpool refresh | Force refresh model cache from API |
How It Works
Multi-Key Rotation
Each OpenRouter key has independent rate limits. ModelPool distributes models across keys:
- - Key1: Model A, Model C, Model E
- Key2: Model B, Model D, Model F
When Key1 hits rate limit → auto-switch to Key2 → back to Key1 when cooldown ends.
2 keys = 2x quota. 3 keys = 3x quota. Simple math.
Smart Fallback Chain
Models ranked by quality (context window × reasoning × benchmarks). Fallback alternates between keys for maximum uptime:
CODEBLOCK1
One-Click Repair (modelpool repair)
7-step auto-fix:
- 1. Check Gateway process
- Test each model API connectivity
- Fix config (
openclaw doctor) - Clean stuck sessions
- Rebuild fallback chain (skip dead models)
- Free memory & clean logs
- Full restart
Getting Started
- 1. Get free API key(s) at openrouter.ai (no credit card needed)
- Run INLINECODE11
- Enter your key(s)
- Done. Go code.
Requirements
- - OpenClaw installed
- Python 3.8+
- At least 1 OpenRouter API key (free)
🎱 ModelPool(免费版)—— OpenClaw 的免费 AI 模型管理器
一条命令。无限免费 AI。再无速率限制。
安装前请准备好你的 OpenRouter API 密钥(在 openrouter.ai 免费获取),然后运行:
bash
modelpool setup
ModelPool 将自动发现最佳免费模型,配置多密钥轮换,并构建智能回退链。你将再也不会遇到速率限制。
命令
| 命令 | 描述 |
|---|
| modelpool setup | 你需要的唯一命令 — 交互式设置,输入密钥,一切自动配置 |
| modelpool auto |
使用现有密钥自动重新配置(非交互式) |
| modelpool list | 浏览所有按质量评分排序的免费模型 |
| modelpool switch
| 手动切换主模型 |
| modelpool status | 显示当前配置、密钥和回退链 |
| modelpool repair | 一键修复 — 自动诊断并修复 7 种常见问题 |
| modelpool keys add | 添加另一个 OpenRouter 密钥(更多密钥 = 更多配额) |
| modelpool keys list | 显示所有已配置的密钥 |
| modelpool refresh | 强制从 API 刷新模型缓存 |
工作原理
多密钥轮换
每个 OpenRouter 密钥都有独立的速率限制。ModelPool 将模型分布到各个密钥上:
- - 密钥1:模型 A、模型 C、模型 E
- 密钥2:模型 B、模型 D、模型 F
当密钥1达到速率限制 → 自动切换到密钥2 → 冷却结束后回到密钥1。
2 个密钥 = 2 倍配额。3 个密钥 = 3 倍配额。简单的数学。
智能回退链
模型按质量排序(上下文窗口 × 推理能力 × 基准测试)。回退在密钥之间交替,以实现最大正常运行时间:
请求 → 密钥1/StepFlash → ✅
请求 → 密钥1/StepFlash → ❌ 速率受限
→ 密钥2/StepFlash → ✅(新密钥!)
→ 密钥1/Nemotron → ✅(不同模型,同一密钥)
一键修复(modelpool repair)
7 步自动修复:
- 1. 检查网关进程
- 测试每个模型 API 连接
- 修复配置(openclaw doctor)
- 清理卡住的会话
- 重建回退链(跳过失效模型)
- 释放内存并清理日志
- 完全重启
快速开始
- 1. 在 openrouter.ai 获取免费 API 密钥(无需信用卡)
- 运行 modelpool setup
- 输入你的密钥
- 完成。开始编码。
要求
- - 已安装 OpenClaw
- Python 3.8+
- 至少 1 个 OpenRouter API 密钥(免费)