What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Most AI clients need three things to use a custom OpenAI-compatible endpoint: its base URL, an API key, and the model ID the endpoint serves. The settings vary by client, and “OpenAI-compatible” does not guarantee support for every API route or feature. The examples below cover ten named clients plus two ways to configure Zed; they are 12 setup patterns, not a definitive roster of 12 different apps.
Before you configure a client
Get the endpoint’s documented base URL, authentication requirements, and model ID from the endpoint provider. Use the URL in the format the provider specifies: some clients or providers expect a base ending in /v1, while others handle route suffixes themselves. Appending /v1 twice can send requests to the wrong path.
As an Amazon Associate I earn from qualifying purchases.
- Confirm the client accepts a custom OpenAI-compatible provider or base URL.
- Enter the documented URL and API key in the client’s provider, connection, or credential settings. A local server may permit a blank key or a specified placeholder; hosted services commonly require an issued key.
- Enter the model ID exactly as the endpoint exposes it. If the client cannot discover models, look for a manual model field or allowlist.
- Test a plain chat request first. Test streaming, tools, image input, embeddings, and other features separately if you need them.
The interface labels and configuration formats below are documented examples, not a guarantee that every product uses the same labels in every release. The available examples do not specify a version date for each client. Check the client’s current documentation if a menu, schema, or field name differs.
12 ways to configure a client or client workflow
The first ten rows name applications. The final two distinguish Zed’s settings-file method from its provider form; they are separate configuration methods, not additional applications.
#1 Best Overall
| Client or setup pattern | Where to configure it | What to supply | Important qualification |
|---|---|---|---|
| Open WebUI | Settings → Admin → Connections → Manage OpenAI API Connections → Add Connection | URL and API key; add model IDs to the allowlist if discovery is unavailable. | Open WebUI’s documentation identifies POST /v1/chat/completions as required for core chat and recommends GET /v1/models for model discovery. |
| Zed provider form | Agent Settings → LLM Providers → Add Provider | Provider name, API URL, model ID, and context window. | Zed’s documented OpenAI-compatible defaults list chat completions and tools as supported, but images and parallel tool calls as unsupported. |
| LibreChat | Configure a custom endpoint in librechat.yaml. |
baseURL and apiKey. |
The integration-matrix example gives this configuration pattern; check LibreChat’s current schema before using it. |
| AnythingLLM | Select “Generic OpenAI” in the provider settings. | base_url and api_key. |
Confirm current field labels in the application. |
| Dify | Select an OpenAI-API-compatible model provider. | base_url and api_key. |
The model-provider workflow may change; check the current interface. |
| Continue | The documented example uses ~/.continue/config.json. |
Provider openai, apiBase, apiKey, and model. |
Continue’s configuration format evolves; verify the current format before copying an older example. |
| Cline | Choose API Provider = “OpenAI Compatible.” | Endpoint URL and key; configure the model using the available model field. | Confirm the model field and current labels in the installed version. |
| Aider | Use the command-line options for the API base and key. | --openai-api-base, --openai-api-key, and --model openai/<model-id>. |
Replace the model ID and connection values with those issued or documented by your endpoint provider. |
| Cursor | Settings → Models → Override OpenAI Base URL. | Base URL and API key. | Check Cursor’s current documentation for plan and model restrictions. |
| n8n | Configure a custom Base URL and credentials in the OpenAI node. | Endpoint base URL and credentials. | Exact steps may depend on the node and n8n version. |
| OpenAI Python SDK-based application | Set the connection values when initializing the SDK client. | base_url, api_key, and the endpoint’s model ID. |
Use the SDK’s current documentation for implementation details; the endpoint must support the API operation your code calls. |
| Zed settings-file configuration | Use the language_models.openai_compatible settings structure. |
API URL, available model name, and model details such as context window. | Match the model ID and context limit to the endpoint. Zed’s guide also describes disabling chat completions for a model that only works with the Responses API. |
The client examples above are based on documented guidance for Open WebUI and Zed and an integration matrix for the other listed patterns. The matrix itself advises checking current product documentation for several configurations, so treat those entries as starting points rather than guaranteed current menus.
Check which API routes and features the endpoint supports
Start with core chat and model discovery
For Open WebUI, the documented core-chat route is POST /v1/chat/completions. Its documentation recommends GET /v1/models for model discovery, but discovery is not required if you can enter model IDs manually in the allowlist. Other clients may use different routes or protocols, so follow both the endpoint’s and client’s instructions.
Rank #2
- Used Book in Good Condition
Test advanced features one at a time
A successful connection or plain chat response does not establish that tool calling, streaming, images, embeddings, or a particular context window will work. These features can depend on both the endpoint and the client’s implementation. Open WebUI’s compatibility guide describes a Gemini case where streamed tool calls do not match the expected OpenAI streaming schema, so native tool calling may fail. Zed’s documented default capabilities also distinguish tools and chat completions from image input and parallel tool calls.
Embeddings are relevant when using retrieval-augmented generation (RAG); image and audio routes enable their respective input or output features. Confirm that the endpoint implements the route and format the client expects before relying on any of them.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Reach local servers from containers and other clients
Use a hostname that resolves from the client’s own environment, not just from a browser on the host machine. When Open WebUI runs in Docker and the model server runs on the host, Open WebUI’s guide says to use host.docker.internal instead of localhost.
For LM Studio connected to Open WebUI, the documented connection URL is http://localhost:1234/v1. Start LM Studio’s Local Server before connecting. The LM Studio documentation says the key may be blank or the placeholder lm-studio; use the value appropriate to the server configuration.
Quick Recap
Best Value
Troubleshoot a connection that fails or only partly works
- Check the URL and route. Compare the base URL with the endpoint provider’s instructions, including whether the client expects the version prefix. Remove any duplicated route suffix.
- Test network access from the client. A URL that works on the host may not resolve from a Docker container or another machine. Check the hostname, port, firewall, and any required browser CORS configuration.
- Check model discovery. If the client expects
/modelsand the endpoint does not expose it in the expected format, enter the endpoint’s model ID manually where the client supports that. - Try plain chat before tools. If chat works but tools fail, check the endpoint’s tool-call format and streaming behavior against the client’s expectations.
- Confirm the protocol. Find out whether the client is sending Chat Completions or Responses requests and whether the endpoint supports that protocol. A server that accepts one is not necessarily compatible with the other.
- Check feature and context limits. Compare tool, streaming, image, embedding, and context-window support on both sides, then test only the features you intend to use.
- Protect the API key. Do not include secret keys in public configuration examples, screenshots, or shared logs.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
