Quickstart

This guide ends with a running local gateway and at least one model you can use from a coding tool or HTTP client. You need macOS or Linux and at least one Codex or Claude Code subscription or configured API-provider credential.

Let your agent do it

The fastest path is to hand setup to the coding agent already working in your terminal. It will ask which subscriptions or API providers you want to connect, run the appropriate RouteKit workflow, and verify the result.

You still approve interactive prompts and complete provider sign-in when required. The agent should not ask you to paste API keys into chat.

Or do it yourself

1. Install RouteKit

curl -fsSL https://github.com/velum-labs/routekit/releases/download/routekit-latest/install.sh | sh
routekit version

The installer supports macOS and Linux, does not require sudo, and provides its own Node.js runtime when a compatible one is not already available. See installation for package-manager and runtime details.

2. Choose what to connect

routekit setup

The guided setup can connect Codex or Claude Code subscriptions and API routes for OpenAI, Anthropic, OpenRouter, or Amazon Bedrock. It checks each selected route, asks you to choose a live default model, and starts the local gateway.

API-provider credentials must already be available in your shell environment. Subscription routes open the matching sign-in flow.

3. Verify the gateway

routekit status
routekit models list

Expected result: routekit status reports a ready gateway at http://127.0.0.1:8080, and routekit models list prints one or more model IDs such as openai/<model>, codex/<model>, or claude-code/<model>.

4. Use a discovered model

Copy one exact model ID from routekit models list.

routekit codex <provider/model>

You can also send an authenticated HTTP request or install persistent coding-tool settings.

Next steps