[fusion-vision] Failed to publish lightweight usage event: $

Error message

[fusion-vision] Failed to publish lightweight usage event: ${formatError(lastError)}

What it means

The fusion-vision MCP tried multiple times to publish a lightweight usage event to the remote usage sync endpoint and every attempt failed; after retries exhausted it logs the last error once (guarded by a flag) and gives up. Usage telemetry for that event is lost but the MCP continues functioning.

Source

Thrown at packages/core/src/mcp/fusion-vision-mcp.ts:434

      }
      const acknowledgement = rawBody ? parseJson(rawBody) : undefined;
      if (isRecord(acknowledgement) && acknowledgement.applied === false) {
        throw new FusionUsageSyncRejectedError();
      }
      usageSyncFailureLogged = false;
      return;
    } catch (error) {
      lastError = error;
      if (attempt >= fusionUsageSyncMaxAttempts || !usageSyncErrorIsRetryable(error)) {
        break;
      }
      await wait(fusionUsageSyncBaseDelayMs * 2 ** (attempt - 1));
    }
  }

  if (!usageSyncFailureLogged) {
    usageSyncFailureLogged = true;
    console.warn(`[fusion-vision] Failed to publish lightweight usage event: ${formatError(lastError)}`);
  }
}

class FusionUsageSyncHttpError extends Error {
  readonly retryable: boolean;

  constructor(readonly statusCode: number) {
    super(`HTTP ${statusCode}`);
    this.name = "FusionUsageSyncHttpError";
    this.retryable = statusCode === 408 || statusCode === 409 || statusCode === 425 || statusCode === 429 || statusCode >= 500;
  }
}

class FusionUsageSyncRejectedError extends Error {
  readonly retryable = false;

  constructor() {
    super("Usage sync endpoint rejected the event.");

View on GitHub (pinned to 99f24806c6)

Solutions

Check network connectivity and proxy settings from the host to the usage endpoint
Inspect the formatted error for HTTP status (401/403 = credentials, 5xx = upstream outage) and fix accordingly
If telemetry is intentionally blocked, suppress the warning or disable usage sync in config to avoid retry churn

Example fix

// before: no proxy configured, egress blocked
// after: route via corporate proxy
process.env.HTTPS_PROXY = "http://proxy.corp:8080";
process.env.HTTP_PROXY = "http://proxy.corp:8080";

Defensive patterns

Strategy: retry

Validate before calling

await fetch(usageEndpoint, { method: "HEAD" }).catch(() => { throw new Error("Usage endpoint unreachable"); });

Try / catch

try { await publishUsageEvent(event); } catch { /* telemetry is best-effort: log once, never fail the MCP call */ }

Prevention

Ensure network/proxy env vars are set for the host
Allowlist the telemetry endpoint in firewall rules
Treat usage-publish failures as non-fatal and rate-limited

When it happens

Trigger: Network egress blocked to the usage endpoint; endpoint returning 5xx or auth failures across all retry attempts; proxy/DNS misconfiguration in the environment running the MCP.

Common situations: Corporate proxies blocking the telemetry host; expired credentials for the usage service; transient outage longer than the retry budget (exponential backoff exhausted).

Related errors

AI-assisted analysis of musistudio/claude-code-router@99f24806c6 (2026-08-27). Data as JSON: /api/errors/80230fb6117c22ef. Report an issue: GitHub.