MCP server
MCP troubleshooting
First check whether the agent runs inside Plous or as a separate MCP client.
Problem
Action
Client cannot find Plous
Check its PATH. Use command -v plous in the same environment. Supply that path only if the client cannot resolve plous .
Server exits at startup
Read stderr. Check the kubeconfig, required context, namespace, and writable data directory.
Connected, but reads fail
Read the Kubernetes authentication, network, or RBAC error. Client sign-in does not grant cluster access.
Wrong cluster
Correct --context and reconnect. Changing the Plous UI does not redirect a standalone server.
Diagnostic tools missing
Enable the relevant diagnostic option and reconnect. Check Kubernetes permissions.
No tools for opening Plous views
These are available only inside Plous.
Earlier evidence missing
Each standalone process has its own evidence profile. Earlier sessions are not searched automatically.
Tool access disabled
Fix the reported storage, history, or target error before reconnecting.
Command timed out
Check whether it ran before retrying. A container process may still be running.
Cleanup pending
Check the reported Job and Pod identities. Deletion is not confirmed yet.
Setup refuses an existing entry
Resolve the unrelated skill or unmanaged plous_debug configuration. Setup does not overwrite it.
For native CLI startup and sign-in, see assistant panel troubleshooting. For profile recovery, see Troubleshooting.