Fresh 2026

Building with coding agents

Configure Cursor, Zed, Continue.dev, OpenCode, Claude Code, and Codex to send requests through Gateway.

Most coding tools, from OpenAI-compatible IDEs like Cursor and Zed to terminal agents like OpenCode, Claude Code, and Codex, let you plug in a custom provider. Point that provider at Merge Gateway and you get every model Gateway supports, including OpenAI, Anthropic, Google, and Kimi, through a single API key. Routing policies, cost governance, and full request observability come built in, visible in the Gateway dashboard.

The three inputs

The OpenAI-compatible IDE guides on this page (Cursor, Zed, Continue.dev) all ask for the same three values:

Input	Value
Base URL	https://api-gateway.merge.dev/v1/openai
API key	Your key from gateway.merge.dev/settings/api-keys
Model name	provider/model format. For example, anthropic/claude-opus-4-6, openai/gpt-5.2, or google/gemini-2.5-pro.

Get the full list of available models from GET /v1/models or the models page in the dashboard.

OpenCode, Claude Code, and Codex follow different setup shapes (a custom provider in opencode.json, an ANTHROPIC_BASE_URL override, and a model provider + profile in ~/.codex/config.toml respectively); see their dedicated guides below for the exact config.

Supported IDEs

Override OpenAI Base URL in Cursor Settings. Supports Ask and Plan modes.

Add Gateway as an OpenAI-compatible provider in settings.json

One-entry config.yaml change. Works in VS Code and JetBrains.

Add Merge Gateway as a custom provider in opencode.json. Works for tool calls and file edits.

Override ANTHROPIC_BASE_URL and auth token. Works for tool calls and file edits.

Register Merge Gateway in ~/.codex/config.toml and opt in via codex --profile merge_gateway

Next steps

Use per-project API keys so IDE usage is tracked and billed separately

Set spend limits and model allow-lists for IDE traffic