← Pricing & setup guides

Roo Code API pricing: control the bill without changing your editor

Your Roo Code model provider determines the inference bill. Compare billing and native tool support together before moving a working setup.

The buying decision

A cheap chat endpoint is useful to Roo only if it supports the native tool calls your agent needs. Test a file edit and its verification before comparing monthly costs.

Keep the agent and change the model account

Roo's OpenAI Compatible settings accept a provider URL, API key and model ID. That lets you evaluate another API without replacing your editor or moving your repository. Record your existing settings first so you can restore the working provider. Provider choice affects model access and inference charges; it does not establish that unrelated hosted services or team products are included.

Native tools are a buying requirement

Roo's current documentation says native tool calling is its only supported tool protocol. A gateway returning fluent chat text has not passed that requirement. Verify that a tool request arrives with a valid name and arguments, the file operation completes, and Roo can use the result in the next turn. Check streaming behavior as well if your workflow depends on incremental tool arguments.

Set up Standard Compute in Roo

Use the extension's OpenAI Compatible provider. Keep your real API key in your local client.

  1. Choose OpenAI Compatible in Roo settings.
  2. Enter https://api.stdcmpt.com/v1 as Base URL and your Standard Compute key as API Key.
  3. Use standardcompute as the model ID.
  4. Match the context and output settings to the model behavior you have selected in the provider dashboard; do not invent token prices for a routed model.
  5. Try one small file change, inspect the diff and run its verification before moving a long task.

Compare three completed tasks

Use a bug fix, a bounded refactor and a test-writing task from the same repository revision. Include retries and failed attempts in the budget consumed. A cheaper token rate can lose its advantage when the agent requires more repair turns. Separate your monthly subscription fee from this sample's allowance consumption; three successful tasks do not establish how much a full working month will cost.

Run your own comparison

Use the same task, starting revision and acceptance check. Record every retry and failed attempt so the result reflects the work you actually paid for.

Sources and setup references