Cowork: 'RPC Failed' Error When Calling a Connector — Fix

You're mid-chat when Cowork tries to call a connector (Slack, Gmail, Drive, Calendar, or any MCP connector) and a red banner appears reading RPC failed or connector tool returned an error. The call is being rejected before any work happens, and the chat stops. The fix is almost always one of three things, in this order: re-authorize the connector, check the upstream service status, or reinstall the connector. All three take under five minutes.

Quick Fix

1. Open the Customize menu in the Cowork sidebar.
2. Click Connectors → find the connector that errored.
3. Click the connector name to view its details.
4. Click Re-authenticate (or Reconnect).
5. Walk through the OAuth flow (or credential entry) in your browser.
6. Return to Cowork and retry the same prompt that failed.

If re-auth succeeds, you're done. If not, continue to the causes below.

Token Expiration (Most Common)

How ScaleIT handles this: See AI Quick-Start

OAuth-backed connectors (Gmail, Drive, Slack, Calendar, Microsoft 365) have tokens that expire on a schedule the vendor controls. When a token expires, the next call to that connector returns an RPC error. Re-authenticating generates a fresh token, and the next call succeeds.

Fix: Follow the Quick Fix steps above. Re-authorization takes about 30 seconds and resolves the majority of RPC errors.

Upstream Service Outage

A real outage on Slack, Google, or Microsoft's infrastructure surfaces as an RPC error in Cowork because the underlying API is unreachable. The connector isn't broken; the vendor's servers are.

Fix: Check the vendor's status page:

• Gmail / Drive: Google Workspace Status
• Slack: Slack Status
• Microsoft 365: Microsoft Service Health

If you see a red incident, wait 10 to 15 minutes and retry. If the status page shows green, move to the next cause.

Corrupted Connector Installation

Rarely, the local connector binary drifts or becomes corrupted, and a re-auth alone doesn't fix the issue. The connector wiring is stale.

Fix:

1. Open Customize → Connectors.
2. Find the broken connector and click the menu icon (three dots).
3. Click Remove.
4. Click Add Connector → search for and add the connector again.
5. Complete the auth flow for the fresh install.
6. Retry the prompt that failed.

A fresh install rebuilds the MCP wiring cleanly and resolves rare cases where the local binary has drifted.

Verify

• The same prompt that produced the RPC error now completes without a red banner.
• The connector appears in the sidebar with a green checkmark or active status indicator.

Still Broken?

If all three steps above are complete and you still see an RPC error:

1. Copy the exact error message from the red banner (e.g., RPC error: tool 'slack_send_message' failed with status 401).
2. Search the Cowork support documentation for the specific tool name and error code.
3. If the error message names a specific tool or rate limit, check that tool's vendor documentation for configuration or quota limits.
4. Contact ScaleIt with the exact error string, the name of the connector, and the steps you've taken. We can investigate whether it's a configuration issue or escalate to the vendor if needed.

Have you been troubleshooting RPC errors across your connectors instead of building workflows? ScaleIt specializes in connector configuration and MCP setup. Contact us for a free consultation.