Open WebUI, Cherry Studio and Chatbox: add a custom OpenAI API

Open WebUI, Cherry Studio and Chatbox are self-hosted or desktop chat apps that let you talk to any model through an OpenAI-compatible API. All three can point at apillms with the same base URL and key, and you pick the model from a dropdown once the connection is saved.

This guide walks through adding apillms as a provider in each app, using the steps from the chat apps guide at /integrations?app=chat. The same guide also covers LobeChat, which follows the identical pattern described below.

What you need

• An apillms account. Sign-up gives 50 free credits to start, and credits never expire.
• Your API key from the API keys page. There's one key per account - use Show to copy it, or Roll key if you ever need to replace it. Rolling disables the old key immediately.
• The OpenAI-compatible base URL: https://apillms.com/v1

That's it. No org ID, no separate project setup, no per-model API key. All three apps below use the same base URL and the same key - you only enter them once per app.

Open WebUI

Open WebUI supports OpenAI-compatible connections in its admin settings. If you're running it yourself (Docker or local install), log in as an admin first.

1. Open Settings → Admin Settings → Connections.
2. Under "OpenAI API", click to add a new connection (some versions show a single OpenAI API URL/key field instead of a list - either works the same way).
3. Set the API base URL to https://apillms.com/v1.
4. Paste your apillms API key into the API key field.
5. Save, then click the refresh icon next to the connection to pull the model list.
6. Open a new chat and check the model picker at the top - the apillms models should now be listed.

You can select a model per chat, or set a default model for new chats in Settings → Interface (or the equivalent default-model field, depending on version). If the connection test fails, Open WebUI usually shows the raw HTTP error from the request - a 401 means the key wasn't saved correctly, and a timeout usually means the base URL is missing or has a typo.

Cherry Studio

Cherry Studio manages providers under its model settings, and lets you keep several providers configured side by side.

1. Go to Settings → Model Providers and add a new provider, or edit an existing OpenAI-compatible entry if you already have one.
2. Set the provider type to OpenAI-compatible.
3. Enter https://apillms.com/v1 as the API host.
4. Enter your apillms API key in the API key field.
5. Click "Fetch models" (or similar) to pull the live model list, then enable the ones you want to appear in the assistant picker.
6. Create or edit an assistant and set its model to one of the apillms models you enabled.

Cherry Studio also lets you set a system prompt and temperature per assistant - these are sent as part of the request body and work the same way they would against any OpenAI-compatible endpoint, so there's nothing apillms-specific to configure there.

Chatbox

Chatbox has a similar provider settings screen, aimed at a quick single-provider setup.

1. Open Settings → Model Provider and choose "Custom Provider" (or the OpenAI-compatible option, depending on the Chatbox version).
2. Set the API host to https://apillms.com/v1.
3. Paste your apillms API key.
4. Click "Load models" or type a model id manually if the list doesn't populate automatically.
5. Select the model you want to chat with and start a new conversation to confirm it responds.

If Chatbox only shows a single "model name" text field instead of a dropdown, type the exact model id from /models (for example claude-sonnet-5 or gpt-5.5) - it will be sent as-is to the /v1/chat/completions endpoint.

Streaming and basic settings

All three apps default to streamed responses when the provider supports it, and apillms' OpenAI-compatible endpoint supports streaming the same way OpenAI's API does, so there's no extra toggle to flip. If you want to test a model's behavior with different system prompts, temperature or max output tokens before wiring it into one of these apps, the console's Playground (after logging in at /login) lets you try the same options and even grab a "Code" snippet of the exact request.

Which model to pick

The full list of on-sale model ids is on /models, with capability filters so you can match a model to the chat app's feature set (vision, tool calling, etc). A few starting points:

For a side-by-side on choosing between Claude, GPT, DeepSeek and Grok for coding work, see Claude vs GPT vs DeepSeek vs Grok. If you also use Claude Code or Claude Desktop alongside one of these chat apps, those connect over the Anthropic-compatible base URL instead - see Connect Claude Desktop to apillms or Set Up Claude Code with apillms.

What a chat session costs

These apps are used for back-and-forth chat, not long agent runs, so the per-message cost is what matters. Using an example message of 1,000 input tokens and 500 output tokens:

• claude-sonnet-5: $0.002 per message, $2.24 for 1,000 messages
• claude-opus-5-5: $0.004 per message, $4.47 for 1,000 messages
• gpt-5.5: $0.004 per message, $4.26 for 1,000 messages
• deepseek-v4-flash: $0.14 for 1,000 messages

Prompt caching is on automatically on apillms, so if your chat app sends a repeated system prompt or long context, the repeated part is billed at the lower cache-read price shown on /pricing instead of the full input price. This matters more for assistants with a long fixed system prompt in Cherry Studio or Open WebUI than for one-off questions.

Since credits never expire and 1 credit = $0.005, a Starter pack ($5 = 1,000 credits) is enough for well over a thousand chat messages on claude-sonnet-5 before you'd need to top up again - check exact per-pack token estimates on /pricing.

Comparing to OpenRouter

If you've used OpenRouter with one of these apps before, switching the base URL and key to apillms is the only change needed - the OpenAI-compatible endpoint shape is the same. List prices on OpenRouter for the same models are higher, for example claude-sonnet-5 is listed at $2 / $10 per 1M input/output tokens there versus apillms' $0.274 / $1.37, and gpt-5.5 is $5 / $30 on OpenRouter versus apillms' $0.532 / $3.19. See a live breakdown at /compare.

Keeping track of usage across apps

Because Open WebUI, Cherry Studio and Chatbox all hit the same account, requests from all three show up together in the console's Usage page, with filters by type and status and an Export CSV option if you want to reconcile spend per app or per model later. The Overview page after logging in also shows balance, spend today and spend over the last 7 days at a glance.

Troubleshooting

• Model list is empty: double-check the base URL ends in /v1 and the key was saved before fetching models. Some apps cache a failed fetch - try removing and re-adding the provider.
• 401 or authentication error: re-copy the key from /keys - if you rolled the key recently, the old one stops working immediately and every app using it needs the new key pasted in.
• Model responds but ignores system prompt or temperature: this is usually an app setting, not an apillms issue - check the assistant/model config in Cherry Studio or Chatbox, since some versions apply these per-assistant rather than globally.
• Request times out or hangs: confirm the app is actually using the OpenAI-compatible endpoint (not an Anthropic-specific one) - apillms' Anthropic-compatible base URL (https://apillms.com, no /v1) will not work with these three apps.
• Need to check spend: the Billing page shows payment history, and your balance is visible from the console after you log in at /login.

Takeaway

Open WebUI, Cherry Studio and Chatbox all connect to apillms the same way: set the base URL to https://apillms.com/v1, paste your key, refresh the model list. Once connected, swap models freely from /models without touching any other setting, and keep an eye on combined usage across all three apps from the Usage page.

FAQ

Do I need a different base URL for each app?

No. Open WebUI, Cherry Studio and Chatbox all use the same OpenAI-compatible base URL, https://apillms.com/v1, with your one apillms API key.

Can I use the Anthropic-compatible endpoint instead?

These three apps expect an OpenAI-compatible provider, so use https://apillms.com/v1. The Anthropic-compatible base URL (https://apillms.com, no /v1) is for apps built against Anthropic's SDK, like Claude Code or Claude Desktop.

Which model should I pick for everyday chat?

claude-sonnet-5 or gpt-5.5 cover most general chat well; deepseek-v4-flash is the cheapest option if you're chatting a lot. Full list and filters are on /models.

Will I be charged if a request fails?

No, failed requests are not billed on apillms.