Cline and Roo Relay Key Setup: Streaming Notes

§1 Connect Cline and Roo with one relay key

Cline and Roo Code can connect to a relay through their OpenAI-compatible provider settings. The basic configuration is straightforward: supply a base URL, a relay API key, and a model identifier. The important details are making sure the endpoint is correct, responses stream incrementally, and the selected model supports the tool interactions your coding workflow needs.

Fluxio fits this setup as a streaming-first, pay-per-token AI API relay. One key provides access to multiple model providers, with replies arriving token by token. For OpenAI-style clients, its API base is https://api.fluxioapi.com/v1.

§2 Step 1: Prepare your relay credentials

Obtain a Fluxio API key and check that your account has available credit. Keep the key out of source files, shared configuration exports, screenshots, and version control. Enter it only in the extension’s designated credential field.

• – Base URL: https://api.fluxioapi.com/v1
• – API key: Your Fluxio relay key, not a direct provider key.
• – Model ID: The exact identifier listed by the relay for your chosen model.

A display name is not necessarily an API identifier. If you choose Claude Haiku 4.5 or Gemini 2.5 Flash, copy the corresponding identifier from the relay’s current model listing rather than guessing its spelling.

§3 Step 2: Configure Cline

Open Cline’s settings and locate the API provider configuration. Choose the option for an OpenAI-compatible endpoint rather than a provider-specific integration. Labels and field placement can vary by extension version.

Enter the Fluxio base URL, paste your relay key, and set the model ID. If Cline exposes separate configurations for planning and acting, configure both deliberately; changing one does not necessarily update the other.

If the extension asks for model capabilities or limits, use the relay’s documented values for that exact model. Do not assume that image input, context size, or tool support carries over from another model. Save the configuration before starting a new task.

§4 Step 3: Configure Roo Code

Open Roo Code’s provider settings and select its OpenAI-compatible option. Add the same base URL and relay key, then enter the exact model ID. Save the configuration profile if your version supports profiles.

Check which configuration is active for the current mode. A successful connection in one profile does not prove that every mode uses the same endpoint. For the first comparison, use the same model in Cline and Roo so that model differences do not obscure configuration problems.

§5 Step 4: Verify streaming before running tools

Start with a small, text-only request, such as asking for a three-sentence explanation of a function. Watch whether the answer appears progressively instead of arriving as one completed block. Some interface buffering is normal, so visible chunk sizes may vary.

If a streaming control is available, enable it. Many coding extensions request streaming automatically, so the absence of a toggle does not mean streaming is unavailable. At the protocol level, an OpenAI-compatible streaming response typically uses server-sent events.

Next, test a harmless workspace action, such as reading a small file with your approval. Streaming text alone does not verify tool compatibility. Confirm that tool requests, results, and the assistant’s follow-up all complete correctly before allowing edits or commands.

§6 Step 5: Troubleshoot the right layer

• – Authentication errors: Check the key, account credit, and accidental whitespace.
• – Endpoint errors: Confirm that the final URL contains one /v1 segment, not a duplicated path.
• – Model errors: Verify the exact identifier and your account’s access to it.
• – Delayed output: Check VPNs, proxies, and gateways that may buffer streaming responses.
• – Broken tool interactions: Verify model support and update the extension before changing unrelated settings.

Review logs carefully, redact credentials before sharing them, and change one setting at a time. Fluxio is useful when you want a shared relay configuration across coding tools, but reliable agent behavior still depends on matching the endpoint, model capabilities, and extension settings.

Sign up for Fluxio.