Official Verified

astrai-inference-router

Route all LLM calls through Astrai for 40%+ cost savings with intelligent routing and privacy controls

skill-install — Terminal

Install via CLI (Recommended)

clawhub install openclaw/skills/skills/beee003/astrai-inference-router

Download Source Code (.zip)

Astrai Inference Router

Route every LLM call through Astrai's intelligent router. Save 40%+ on API costs. Privacy controls built in.

What it does

Smart routing: Classifies each task (code, research, chat, creative) and picks the optimal model
Cost savings: Bayesian learning finds the cheapest provider that meets your quality threshold
Auto-failover: Circuit breaker switches providers when one goes down
PII protection: Personally identifiable information stripped before reaching any provider
EU routing: GDPR-compliant European-only routing with one setting
Budget caps: Set daily spend limits to prevent runaway costs
Real-time tracking: See exactly how much you're saving per request

Setup

Get a free API key at as-trai.com
Set ASTRAI_API_KEY in your environment or skill config
Choose your privacy mode (default: enhanced)
Done — all LLM calls now route through Astrai

Privacy Modes

standard: Full routing intelligence, normal logging
enhanced: PII stripped, metadata-only logging, region enforced
max: Zero data retention, EU-only, all PII stripped, no prompt logging

Environment Variables

Variable	Required	Description	Default
`ASTRAI_API_KEY`	Yes	Your API key from as-trai.com	—
`PRIVACY_MODE`	No	standard, enhanced, max	enhanced
`REGION`	No	any, eu, us	any
`DAILY_BUDGET`	No	Max daily spend in USD (0 = unlimited)	10

External Endpoints

Endpoint	Purpose	Data Sent
`https://as-trai.com/v1/chat/completions`	LLM inference routing	Prompts (with PII stripped if enhanced/max mode)
`https://as-trai.com/v1/signup`	Free API key registration	Email address

Security & Privacy

All requests authenticated via API key in Authorization header
PII stripping runs locally before any data leaves your machine (enhanced/max modes)
EU routing mode ensures prompts never leave European infrastructure
Zero data retention available in max privacy mode
No credentials are stored by the skill — only your API key in environment variables
Source code is fully open: github.com/beee003/astrai-openclaw

Model Invocation

This skill intercepts outgoing LLM API calls and reroutes them through the Astrai gateway. The gateway selects the optimal provider and model based on task type, cost, and quality. Your prompts are processed by third-party LLM providers (Anthropic, OpenAI, Google, Mistral, etc.) according to your region and privacy settings.

Pricing

Free: 1,000 requests/day, smart routing, failover
Pro ($49/mo): Unlimited requests, EU routing, PII stripping, analytics
Business ($199/mo): Multi-agent dashboards, compliance exports, SLA

Read Full Documentation on GitHub

Metadata

Author@beee003

Stars2387

Updated2026-03-09

View Author Profile

AI Skill Finder

Not sure this is the right skill?

Describe what you want to build — we'll match you to the best skill from 16,000+ options.

Find the right skill

Add to Configuration

Paste this into your clawhub.json to enable this plugin.

{
  "plugins": {
    "official-beee003-astrai-inference-router": {
      "enabled": true,
      "auto_update": true
    }
  }
}

Related Skills

sealvera

Tamper-evident audit trail for AI agent decisions. Use when logging LLM decisions, setting up AI compliance, auditing agents for EU AI Act, HIPAA, GDPR or SOC 2, or when a user asks about AI decision audit trails, explainability, or SealVera.

ahessami123 2387

model-fallback

Multi-model automatic fallback system. Monitors model availability and automatically falls back to backup models when the primary model fails. Supports MiniMax, Kimi, Zhipu and other OpenAI-compatible APIs. Use when: (1) Primary model API is unavailable, (2) Model response time is too slow, (3) Rate limit exceeded, (4) Need to optimize costs by using cheaper models for simple tasks.

azure5100 2387

semantic-router

让 AI 代理根据对话内容自动选择最合适的模型。四层识别（系统过滤→关键词→指示词→语义相似度），四池架构（高速/智能/人文/代理），五分支路由，全自动 Fallback 回路。支持 trigger_groups_all 非连续词组命中。

halfmoon82 2387

subagent-isolation-guard

固化子代理物理隔离与语义路由旁路。防止跨代理上下文污染及由于语义路由导致的子代理切模/重置问题。