Skip to navigation
DeploymentRoll out to employees

Roll out models

Route every employee's AI client through Gateway

To route an AI client through Gateway, replace its model provider settings with three values:

SettingValue
Base URLhttps://api-gateway.merge.dev/v1/anthropic for Claude clients, https://api-gateway.merge.dev/v1/openai for OpenAI-compatible clients
API keyThe employee’s own Gateway key
Modeldefault_routing, so your routing policy picks the model

Give each employee their own key

Use a key per employee, so spend is attributed to them and their Group’s budget applies. Employees get one of two ways:

  • Workforce CLI: mfw login signs them in and creates a key for the machine
  • Workforce console: Get started → Set it up myself shows a personal key and the setup steps for their client

Avoid one shared key. It hides who spent what and has to be rotated everywhere if it leaks.

Supported clients

Claude Code, Codex, Claude Desktop, Zed, Continue.dev, OpenCode, Pi, Factory Droid, and Cursor.

Cursor routes only its Ask and Plan modes. Its agent mode and Tab completion always use Cursor’s own models.

Confirm employees are routing

Run a pilot with a few employees first. Their requests appear in LLM calls within seconds. Anyone on your employee list with no requests isn’t routing yet.