Errors you may see in your AI tool
session_expired or “Please re-authenticate via /mcp”
Your Condense sign-in has expired. During the beta a sign-in lasts 30 minutes. Run your AI tool’s sign-in step again:
- Claude Code: type
/mcp. - Gemini CLI: run
/mcp auth condense. - Codex: run
codex mcp login condense.
Access denied
The tool is disabled for your account or role. Ask your Condense administrator to enable it.Messages on the approval page
- No Environments Assigned. Ask your administrator to assign you an environment, then start again from your AI tool.
- Authentication Session Expired. Return to your AI tool and start the connection again. The access link works for about 5 minutes.
- Authentication Failed. Close the tab and try connecting your AI tool again.
Troubleshooting
Select a symptom to see the likely cause and what to do.No Condense tools appear
No Condense tools appear
Likely cause: The sign-in did not finish, the URL or field name is wrong for your AI tool, or no environment is assigned to you.What to do: Follow If no tools appear.
It worked a moment ago and now says the sign-in expired
It worked a moment ago and now says the sign-in expired
Likely cause: During the beta a sign-in lasts 30 minutes.What to do: Sign in again from your AI tool.
Reconnecting does not work
Reconnecting does not work
Likely cause: An administrator revoked your access or paused the connection.What to do: Ask your administrator to restore your access, then reconnect.
A tool shows Expired on the AI Tools page
A tool shows Expired on the AI Tools page
Likely cause: The sign-in for that connection expired. During the beta a sign-in lasts 30 minutes.What to do: Sign in again from your AI tool. See Verify and reconnect.
A tool group shows Disabled by Admin
A tool group shows Disabled by Admin
Likely cause: An administrator switched it off in this environment.What to do: Ask your administrator to enable it. You cannot turn it on yourself.
A workspace you expect is missing from the list
A workspace you expect is missing from the list
Likely cause: You are not authorized for that workspace in the environment you chose.What to do: Ask your administrator for the workspace role. You only see workspaces you are authorized for.
Your AI tool reads the wrong environment
Your AI tool reads the wrong environment
Likely cause: A connection reads the one environment you chose when you approved it.What to do: Connect again and choose the environment you meant.
Your AI tool cannot get a pod status, pod metrics or job details
Your AI tool cannot get a pod status, pod metrics or job details
Likely cause: These tools need a deployment or job name and its namespace, and one of them is missing.What to do: Name both in your request, for example “the status of the deployment fleet-api in the namespace fleet-tracking”.
A metrics query returns nothing
A metrics query returns nothing
Likely cause:
Execute-query runs a PromQL query, and the query may name a metric that does not exist.What to do: Ask your AI tool to list the available metrics first with List-Metrics, then query one of them.Connector logs come back empty or very short
Connector logs come back empty or very short
Likely cause: Logs are collected for a duration in milliseconds, and the default is 200.What to do: Ask for a longer duration, such as “the last 5000 ms”.
A tool call is refused because of a limit
A tool call is refused because of a limit
Likely cause: Your organization reached the 5-hour limit or the daily limit. Everyone in the organization shares them. The limits can change during the beta.What to do: Wait for the limit to reset. It resets every 5 hours or every day. The AI Tools page shows your organization’s usage for the last 5 hours and for today. See Usage and limits.
A deploy is refused
A deploy is refused
Likely cause: Your role does not allow deploying in that workspace, or the environment is in Strict read-only mode.What to do: Ask your administrator. See Scope and permissions.