Radicalbit Ai Gateway

Radicalbit AI Gateway — Claude Code Plugin

Official Claude Code plugin from Radicalbit for configuring and managing the Radicalbit AI Gateway.

Skills

Skill	Description
/radicalbit-ai-gateway:ai-gateway-config	Generate or update the AI Gateway config.yaml

Installation

Via Claude Code Marketplace

Add the marketplace and install the plugin:

/plugin marketplace add radicalbit/radicalbit-skills
/plugin install radicalbit-ai-gateway@radicalbit

Or browse interactively with /plugin.

Team Installation (project-level)

Add to your project's .claude/settings.json to share with your team:

{
  "extraKnownMarketplaces": {
    "radicalbit": {
      "source": {
        "source": "github",
        "repo": "radicalbit/radicalbit-skills"
      }
    }
  },
  "enabledPlugins": {
    "radicalbit-ai-gateway@radicalbit": true
  }
}

Team members receive installation prompts upon trusting the repository.

Usage

Invoke skills using the /radicalbit-ai-gateway:<skill-name> syntax in Claude Code.

ai-gateway-config

Generates or updates the Radicalbit AI Gateway config.yaml file based on your requirements.

Trigger phrases:

• "generate gateway config"
• "create config.yaml"
• "configure the gateway"
• "add a guardrail"
• "set up caching"
• "add rate limiting"
• "configure fallback"

Example interaction:

You: Generate a gateway config with two OpenAI routes — one for customer support
     with PII anonymization and semantic caching, and one internal QA route
     with keyword routing.

Claude: [Invokes /radicalbit-ai-gateway:ai-gateway-config and writes config.yaml]

What the skill produces:

• A complete config.yaml written to your current working directory
• Routes with models, guardrails, caching, rate/token limits, and fallbacks
• !secret ENV_VAR_NAME placeholders for all API keys (never hardcoded)
• A summary of what was generated and which fields need real values

About the Radicalbit AI Gateway

The Radicalbit AI Gateway is an OpenAI-compatible proxy that provides:

• Multi-provider routing — OpenAI, Anthropic, Ollama, vLLM, and more
• Guardrails — PII anonymization, toxicity detection, prompt injection blocking
• Caching — exact and semantic caching backed by Redis/Valkey
• Rate & token limiting — protect against abuse and control costs
• Fallback chains — automatic model failover for high availability
• Advanced routing — keyword, token-length, time-based, and budget-aware routing
• Observability — Prometheus metrics for Grafana dashboards

License

Licensed under the MIT License.

Contributing

See CONTRIBUTING.md for how to add or update skills.