AISIX AI Gateway — Claude Code MCP server: features and setup
A self-hosted gateway routes language-model requests through one API. Use it to manage provider access and model routing in one place.
Open on GitHubWhat AISIX AI Gateway does
AISIX accepts requests in OpenAI or Anthropic format and forwards them to the configured provider. The description names OpenAI, Anthropic, Google Gemini, AWS Bedrock, Azure OpenAI, and DeepSeek; gateway features include SSE streaming, rate limits, caching, and observability. This is an LLM traffic gateway, not a description of MCP tools.
Run it as a static binary or container. Define resources in resources.yaml and apply changes with SIGHUP without restarting; for a multi-replica cluster, use etcd instead of resources_file. The quickstart uses Docker and the OPENAI_API_KEY and CALLER_API_KEY environment variables. The metrics and status listener on port 9090 has no authentication, so keep it on loopback or a private network.
What it helps with
- route requests to different LLM providers through one API
- restrict model access with caller API keys
- apply resource changes without restarting the gateway
- expose LLM traffic metrics through Prometheus
Example prompts for the agent
- Set up AISIX to route OpenAI requests using
resources.yaml. - Prepare an AISIX configuration with caller credentials and an allowed model.
- Show how to send model requests through the AISIX API.
How to install AISIX AI Gateway
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about AISIX AI Gateway
How do I install AISIX AI Gateway in Claude Code?
The description provides no Claude Code installation command. Its quickstart uses Docker: prepare config.yaml and resources.yaml, set OPENAI_API_KEY and CALLER_API_KEY, then run the ghcr.io/api7/aisix:latest container with both configuration files.
What does AISIX AI Gateway do?
It accepts requests in OpenAI or Anthropic format and routes them to configured LLM providers. It also provides routing controls, rate limits, caching, and observability.
Which environment variables does the AISIX quickstart use?
The Docker example sets OPENAI_API_KEY for the provider credential and CALLER_API_KEY for the caller credential. The metrics and status listener on port 9090 is unauthenticated.
