Agent Setup
Step-by-step configuration for each supported AI coding agent.
User Guide
This guide is for developers using the AI Gateway to route LLM requests from AI coding agents and applications.
Who This Is For
Section titled “Who This Is For”You are a developer who wants to:
- Route AI agent requests through a centralized gateway instead of hitting provider APIs directly
- Use any of the supported AI coding agents (Claude Code, OpenCode, Goose, Continue.dev, LangChain, Codex CLI)
- Access multiple model providers (Bedrock, Anthropic, OpenAI, Google, Azure OpenAI) with a single set of credentials
- Avoid managing individual provider API keys on your local machine
What You Can Do
Section titled “What You Can Do”| Capability | Description |
|---|---|
| Route across providers | The gateway config selects the provider and fails over across a priority chain (Bedrock primary, Anthropic-direct fallback) without client changes |
| Use either API format | Send requests in OpenAI Chat Completions format (/v1/chat/completions) or Anthropic Messages format (/v1/messages) on one port |
| Authenticate once | Use a single Cognito M2M token instead of per-provider API keys |
| Use any compatible agent | Configure Claude Code, OpenCode, Goose, Continue.dev, LangChain, or Codex CLI |
Sections
Section titled “Sections”API Reference
Endpoints, headers, request/response formats, and rate limits.
Routing Strategies
Provider failover and load balancing via agentgateway priority groups.
Troubleshooting
Solutions for common errors: 401s, 403s, missing headers, and more.