At a glance
- Availability: Experimental (how to enable).
- Auth: API key.
- Connection: The key comes from
OPENROUTER_API_KEY. - Docs: https://openrouter.ai/docs/api/reference/overview
Credentials
Set these per environment. See Connect an integration.| Variable | Required | Description |
|---|---|---|
OPENROUTER_API_KEY | Yes | OpenRouter API key (starts with sk-or-) Docs. |
Setup
- Create an OpenRouter account: Go to https://openrouter.ai and sign up or sign in. Some models offer free variants for testing; paid models require purchasing credits under Settings → Credits.
- Create an API key: Open https://openrouter.ai/settings/keys, click ‘Create Key’, and optionally set a credit limit for the key.
- Store the key: Copy the key and add it to your .env file as OPENROUTER_API_KEY=sk-or-…
- Verify access: Run the List Models tool to confirm the key works, then try a chat completion with a model like openai/gpt-4o-mini.
- Model IDs are namespaced as vendor/model, e.g. anthropic/claude-sonnet-4
- Chat completions consume prepaid credits; use the Get Generation tool to inspect the exact cost of a request
- The Get Credits endpoint requires a management key per OpenRouter’s docs - a regular inference key may be rejected there
Tools
| Tool | Access | Description |
|---|---|---|
| List Models | Read | List the models available through OpenRouter |
| Create Chat Completion | Write | Generate a chat completion from any OpenRouter model |
| Get Generation | Read | Get metadata, token counts, and cost for a past generation by ID |
| Get Current Key | Read | Get the current API key’s label, credit usage (total/daily/weekly/monthly in USD), spend limit, remaining limit, and free-tier status - works with the regular inference key |
| Get Credits | Read | Get total credits purchased and used for the account (OpenRouter requires a management key for this endpoint) |
Example prompts
- List the models available through OpenRouter and highlight a few with large context windows.
- Use OpenRouter to send the same prompt to two different models and compare the responses.