--- title: Using models and API keys description: Answers about using models on Frely: API keys, client settings, usage, and what to do when a request fails. applies_to: any updated: 2026-10-26 order: 3 --- # Using models and API keys ## How do I create an API key? Open the API key page in your console and create one. A key authorizes model calls, not management actions. Keep it in your client's protected configuration. Never put it in Git, screenshots or chat. Revoking a key stops later use. Related: [Guides](https://frely.cloud/docs/) Console: [Your API keys](https://frely.cloud/user/account/keys) API tools: `keys.list`, `keys.create`, `keys.set_limits` ## What is the Frely API base URL? https://api.frely.cloud. Always use https: requests over plain http are rejected, so a client configured with http://api.frely.cloud will fail. The same address serves every protocol, for example https://api.frely.cloud/v1/chat/completions, /v1/responses, /v1/messages, /v1/embeddings and the model list at /v1/models. Some clients want the base URL with /v1 on the end (https://api.frely.cloud/v1) and add the endpoint path themselves. Related: [Use Frely in Claude](https://frely.cloud/docs/claude/) · [Use Frely in ChatGPT](https://frely.cloud/docs/chatgpt/) Console: [API test (shows the base URL and a ready curl command)](https://frely.cloud/user/library/api-test) API tools: `models.available` ## How do I authenticate API requests? Send your API key as a bearer token: Authorization: Bearer . Add Content-Type: application/json to requests with a body. A request without a valid key gets a 401. A key authorizes model calls only, not account management. Console: [Your API keys](https://frely.cloud/user/account/keys) API tools: `keys.list`, `keys.create` ## What does an error response look like? A JSON body of the form {"error":{"message","type","param","code"}}. The type follows the HTTP status: 401 authentication_error, 402 insufficient_quota, 403 permission_error, 404 not_found_error, 408 or 504 timeout_error, 429 rate_limit_error, other 4xx invalid_request_error, 503 service_unavailable_error, other 5xx server_error. Every response carries an x-request-id header; keep it when you contact support. A retryable failure may include a Retry-After header in seconds. Related: [Contact support](https://frely.cloud/support/) Console: [Request history](https://frely.cloud/user/usage/requests) API tools: `request_history.list` ## What do I enter in my AI app or client? Three values: the public service address for the protocol you chose (not an internal address), your own API key, and a public model name that your account can see. Your account instructions show the exact values. Related: [Use Frely in Claude](https://frely.cloud/docs/claude/) · [Use Frely in ChatGPT](https://frely.cloud/docs/chatgpt/) API tools: `models.available`, `keys.list` ## How do I add Frely to Claude or ChatGPT? Add Frely as a connector, use the FrelyMCP address as the MCP server URL, choose OAuth, sign in to Frely and approve access. The step-by-step guides show each screen. Related: [FrelyMCP guide](https://frely.cloud/docs/frelymcp/) · [Use Frely in Claude](https://frely.cloud/docs/claude/) · [Use Frely in ChatGPT](https://frely.cloud/docs/chatgpt/) ## A model shows in the list but I cannot call it. Being listed is not permission to call. Check your plan, balance, allowance and the service terms, and ask an authorized administrator if you are on a Team. Rotating keys does not get around a restriction. API tools: `models.available`, `budget.get`, `credits.account` ## My request failed or timed out. Keep a redacted note of the time, protocol, model name and request ID, and look at your request history. A timeout or dropped connection does not prove nothing was used, so check usage before retrying. Do not send credentials, request bodies or raw error dumps to anyone. Related: [Contact support](https://frely.cloud/support/) Console: [Request history](https://frely.cloud/user/usage/requests) API tools: `request_history.list`, `usage.summary` ## Can I rely on model output? Model output can contain errors and Frely does not guarantee accuracy. Review it before using it in safety-critical, legal, financial, medical, security or production settings. Related: [Terms of Service](https://frely.cloud/terms/)