Cline returns 503 or no healthy upstream? Preserve the provider response.

A 503 response may come from a provider or gateway rather than Cline itself. Save the exact response and stop duplicate retries while you check whether the failure is transient or request-specific.

Do this first: Record the selected provider/model, HTTP status, response body, request ID, timestamp, and any Retry-After header. Stop a repeated retry cycle before making another model or configuration change.

What the failure looks like.

Cline reports HTTP 503, API streaming failure, or the literal text no healthy upstream. These strings identify a failed request path, but not which service or hop failed.

Why it happens.

Issue #4205 is a specific Cline report, not proof that every 503 or no-healthy-upstream response originates in Cline. Identify the provider endpoint and compare a single controlled request before changing project state.

Fix it without losing the task.

Copy the complete status, response body, request ID, timestamp, and any Retry-After value. Redact API keys and private prompt content.

Confirm the provider/model and endpoint selected in Cline. Check the provider status using its official channel before changing the project.

After the request is known to have stopped, make at most one controlled retry or endpoint comparison and record the result. Avoid changing provider, model, and settings together.

Inspect git diff and test only the files affected by the failed task. If the error persists, report the redacted request evidence upstream.

What the next AI should receive.

Cline version and surface, provider and model, endpoint type, exact HTTP response, Retry-After/request ID if present, timestamp, retry count, changed files, and one controlled comparison result.

First safe action

Save the response body and request identifiers, then check the selected provider status and configured endpoint. Do not send parallel retries while an earlier request may still be active.

Evidence boundary

A 503 or gateway message alone does not identify the failing component or promise a retry time. Provider-specific limits and status pages vary; this guide is not a universal Cline backend diagnosis.

Diagnostic checkpoint

This page is only for an observed 503/no-healthy-upstream response. For a broader Cline request-stall investigation across error codes, use the Cline API request guide.

How ShardStitch helps.

ShardStitch can preserve the project diff, exact error, failed attempts, and next diagnostic for a new session. It cannot query provider health or prevent Cline from retrying a request.

Source trail.

Cline issue #4205 and the Cline issue tracker.

FAQ.

Does no healthy upstream mean the Cline service is down?

Not necessarily. The message may be returned by a provider or gateway on the request path. Preserve the endpoint and response details, then check the relevant provider status.

Should I keep retrying a Cline 503 response?

Not in a rapid loop. Check whether the previous request ended and follow a returned Retry-After value if present. Record the outcome of one controlled retry.

Scope note: This guide focuses on provider/gateway 503 responses observed in Cline. Issue #4205 is one report; it does not establish prevalence or one universal remediation.