Why does a Cline setup stop working after the provider changes?
Moving an existing Cline setup should not mean rebuilding your workflow from scratch. If your editor previously connected to one provider and you are now pointing it at SwiftRelay, the break usually comes down to three things: the base URL, the API key, or the model selected for the request.
SwiftRelay is a pay-per-token API relay. You top up a balance, and each call deducts only the tokens it used. There is no monthly fee and nothing to subscribe to, so you can connect in minutes without changing how you budget the service around a recurring charge.
For Cline, the important pairing rule is simple: Cline uses the OpenAI-compatible endpoint and works with GPT, DeepSeek and the other remaining models. If the model you selected in Cline belongs to a different client family, the request will not work through this endpoint, even when the base URL and key are correct.
Start by separating configuration from model choice. A valid SwiftRelay connection needs the API host, the OpenAI-compatible base URL and a swft_ key. Once those are in place, choose one of the supported models from the list below and make a small test call. That gives you a clean first run before you bring your larger Cline tasks back into the lap.
How do I fix a “401 Unauthorized” message in Cline?
A “401 Unauthorized” response usually means Cline reached the endpoint but did not present a key the service accepts. Check the API key field first. It should contain your SwiftRelay key, including the swft_ prefix, rather than a key copied from another provider.
Next, check the base URL. For an OpenAI-compatible client such as Cline, use:
https://api.fastapirelay.com/v1
Then place the same swft_ key in the API key field. Do not add the base URL to the key field, and do not leave an old provider’s address in a separate environment variable that overrides the setting inside Cline. If you have changed keys since creating the setup, replace the old value completely so no extra spaces or characters remain at either end.
Your account can hold up to 20 live API keys, which makes it practical to keep one key for Cline and separate keys for other work. The key still needs to be active, and the account needs enough balance for the request to run. SwiftRelay also offers a $0.22 sign-up gift credit, so a new account has a little room for a first test.
After saving the settings, run a short prompt in Cline. If the unauthorized message disappears, the connection is working. If it remains, revisit the exact base URL and key format before changing anything else. This is a short pit stop: verify the two values, make one small call, and check the result.
Why does Cline report “404 model not found” after I switch to SwiftRelay?
A “404 model not found” message points to the model identifier rather than the key. Cline may still be sending the model name from your previous setup. Open the model selection in Cline and replace it with a model that SwiftRelay lists for the OpenAI-compatible endpoint.
For example, you can use DeepSeek V4 Flash with the identifier deepseek-v4-flash, DeepSeek V4 Pro with deepseek-v4-pro, GPT-6 Astra with gpt-6-astra, or GPT-6.1 Sol with gpt-6.1-sol. Copy the identifier exactly as shown, including the hyphen and the version numbers. A display name such as “GPT-6.1 Sol” is not the same thing as the API identifier gpt-6.1-sol.
The model must also match the client route. Cline pairs with GPT, DeepSeek and the other remaining models through the OpenAI-compatible endpoint. Selecting a model from a different family will not produce a clean Cline run, even if the model name looks familiar and the request reaches the API host.
Once the identifier is correct, save the model setting and try again. Usage in SwiftRelay lists the tokens and cost for each call, so you can confirm that the request completed rather than relying only on the editor’s status message. That closes the loop between the Cline error and the SwiftRelay dashboard.
What should I check when Cline shows a connection error?
Connection errors often come from an incomplete base URL or from a value that was pasted with instructions around it. The OpenAI-compatible base URL for Cline is:
https://api.fastapirelay.com/v1
Use that address as the base URL. The API host is https://api.fastapirelay.com, while the OpenAI-compatible clients use the /v1 path. Do not paste a full configuration example into a single field when Cline provides separate fields for the base URL and the key.
Then check the key one more time. It should be only the swft_ key, and it should be entered in the API key field. If your setup imports environment variables, make sure the imported value does not override the value you just saved in the dashboard or client settings.
It also helps to make the first request small. A short prompt gives you a quick signal that the endpoint, key and model all agree. If that succeeds, you can return to the task that prompted the migration. If it fails, the dashboard usage list provides the clearest next clue: no recorded call suggests the request did not reach SwiftRelay, while a recorded call gives you the token count and cost to inspect.
There is no need to subscribe to a recurring bundle to test the connection. Top up a balance, run the call, and let the usage deduct what the request actually used. That makes the first lap inexpensive and easy to measure.
How do I check token pricing and cost per call on SwiftRelay?
Because SwiftRelay bills per token, the model you choose directly affects what a Cline session costs. The current prices per 1M tokens on the SwiftRelay pricing page are:
- DeepSeek V4 Flash (deepseek-v4-flash): input $0.2639, output $1.06, cache read $0.006944.
- DeepSeek V4 Pro (deepseek-v4-pro): input $1.17, output $3.51, cache read $0.0389.
- GPT-6 Astra (gpt-6-astra): input $0.6597, output $3.28, cache read $0.0653.
- GPT-6.1 Sol (gpt-6.1-sol): input $0.1736, output $0.8542, cache read $0.0347.
To make these numbers concrete, here is an example using the listed prices. If a call used 1M input tokens and 1M output tokens with GPT-6.1 Sol, the token cost would be $0.1736 + $0.8542 = $1.0278. A call with the same token volumes using DeepSeek V4 Flash would cost $0.2639 + $1.06 = $1.3239.
These are examples for one million input tokens and one million output tokens, not a prediction of what every Cline task will use. A shorter request will use fewer tokens. The usage list in SwiftRelay shows the tokens and cost for each call, which lets you compare actual sessions instead of guessing.
Available top-up amounts are $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. Start with an amount that suits your workflow, run your Cline calls, and watch the balance move by the amount each request used. There is no monthly fee and nothing to subscribe to.
What happens if I need a refund or help with the migration?
If a top-up does not suit your needs, SwiftRelay provides a refund for unused credit within 11 days of the top-up, minus what was used and the payment fee. Review your usage before deciding, and remember that the refund applies to the unused portion rather than the whole top-up.
SwiftRelay also has a referral reward structure: you receive 14% of each top-up made by someone you invited, and 3% from the people they invite. Rewards can be withdrawn with a minimum withdrawal of $4.86, or moved into your balance. The balance option can be useful when you want the credit available for the next Cline run.
For support, use Messages in the dashboard. Support is available there for signed-in users, and there is no support email. When you write a message, include the error text, the model identifier, the base URL you used and whether SwiftRelay recorded a call in the usage list. Those details turn a vague connection problem into a focused check.
Your migration checklist is now short: top up, create a swft_ key, set the OpenAI-compatible base URL, select a GPT or DeepSeek model identifier, and make one small test call. The stopwatch is ready, the first lap is clean, and SwiftRelay is ready when you are — sign up now and move your Cline setup over in minutes.