Edy Cloud AI

Editor integration

Edy Cloud AI for VS Code

Edy Cloud AI registers Edy as a native VS Code language-model provider with secure browser sign-in, account-aware models, and streaming tool support.

Install and connect

  1. 01

    Install the extension

    Install Edy Cloud AI from the VS Code Marketplace.
  2. 02

    Sign in from VS Code

    Open the Command Palette and run Edy: Sign In. The extension opens Edy in your system browser without asking you to copy a credential.
  3. 03

    Sign up or log in

    Create an Edy account or log in on the website, then choose Open Visual Studio Code. Edy returns a short-lived, single-use authorization code protected by PKCE; access and refresh tokens never appear in the browser URL.
  4. 04

    Choose a model and code

    Back in VS Code, open Chat and select an Edy Cloud model. The picker shows the models currently available to your account.

Models, free mode, and paid mode

The extension sends the selected public model slug to Edy; provider IDs remain internal. Free models use the shared Free-model allowance without wallet billing in either account mode. Premium models require Paid mode, use the separate Paid allowance, and bill the wallet at their configured token prices.

Browse models Understand billing

Agent tasks and free-model limits

VS Code may call the selected model several times during one agent task as it reads files, receives tool results, and prepares the final response. Each model call contributes to the model class's RPM, TPM, and daily counters.

If a free-model burst limit is reached, the extension shows the reset delay returned by Edy. The request is not moved to Paid limits and does not debit the wallet. You can inspect the exact quota reason and Edy request ID in the Edy Cloud Output channel.

How agent requests are counted

Troubleshooting and request IDs

Open View → Output and select Edy Cloud. Every response logs Edy's request ID and HTTP status. Include that Edy request ID when reporting a failed request; it is different from VS Code's Client Request Id.

Temporary provider overload

Edy retries HTTP 500, 502, 503, 504, and 529 failures up to twice, but only before response output starts. If all attempts fail, the stream returns upstream_overloaded. Wait briefly and retry; your unused billing reservation is released.

Rate limit reached

HTTP 429 identifies whether the request hit RPM, TPM, daily requests, or daily tokens and includes a retry delay. Start a new attempt after the stated reset; repeated retries before then remain blocked.