Cursor IDE Integration

Overview

Cursor supports OpenAI-compatible API providers. By pointing Cursor at TheRouter.ai you get access to 200+ models, automatic fallbacks, and unified billing across all providers.

Configuration

Open Cursor Settings and navigate to Models. Add a new OpenAI-compatible provider with these values:

Provider type: OpenAI-compatible
Base URL: https://api.therouter.ai/v1
API key: <your-therouter-api-key>

Replace <your-therouter-api-key> with a tenant key from the TheRouter.ai dashboard. After saving, let Cursor refresh the model list from/v1/models.

Model Selection

Use provider/model format when selecting models in Cursor. Popular choices for coding tasks:

anthropic/claude-sonnet-4.6   — Best default for coding
anthropic/claude-sonnet-4.5   — Stable high-quality fallback
openai/gpt-4o                 — Strong general-purpose model
google/gemini-2.0-flash-001   — Fast large-context option
anthropic/claude-opus-4.6     — Deep reasoning for complex tasks

Validation

Before picking a model in Cursor, verify that your key can list the exact model IDs exposed by TheRouter.ai:

curl https://api.therouter.ai/v1/models \
  -H "Authorization: Bearer $THEROUTER_API_KEY" | head -c 400

Copy model IDs exactly from that response. Do not use upstream provider model names or dated variants unless they appear in /v1/models.

Troubleshooting

If Cursor reports model not found or a generic API error, the cause is usually one of these:

1. The Base URL is missing the /v1 suffix.
2. The API key is a management key instead of a tenant key.
3. The selected model ID does not exactly match an id returned by /v1/models.
4. Cursor cached an old model list; refresh the provider after updating the key.

Start debugging with /v1/models first. If the model is not in that list, Cursor will not be able to use it.