Troubleshooting

Start with anc auth status. Most problems are one of the six access states, and the error text names which one.

Setup and authentication

Symptom	Cause and fix
Anypoint Connect is not configured	No credentials for the active profile. Run anc config init, or export ANYPOINT_CLIENT_ID and ANYPOINT_CLIENT_SECRET. Logging in cannot help yet
Not authenticated. Run: anc auth login	Credentials exist, no usable token. Log in
Token expired and no refresh token	The stored session cannot be renewed. anc auth login again
Token refresh failed	The refresh token was revoked, or the Connected App changed. Log in again; if it recurs, check whether the app was rotated in Access Management
The browser never returns	The Connected App's redirect URI does not match http://localhost:3000/api/callback, or port 3000 is occupied
Commands hit the wrong organization	An exported ANYPOINT_PROFILE or environment variable outranks the profile you expected. anc config show reports the resolved source. See Profiles

Environments and permissions

Symptom	Cause and fix
Environment "X" not found. Available: …	Misspelling, or you are in a different organization or business group. Compare against the printed list before assuming a permission problem
A 403 on one operation while others work	The Connected App lacks that scope. Grant it in Access Management and log in again
MQ or Object Store tools fail consistently	The subscription may not include the feature. Check get_entitlements — a 403 on an unprovisioned service is correct behavior
Audit log returns nothing	The View Audit Logs scope is optional and often ungranted. Grant it, or accept the gap

Logs and metrics

Symptom	Cause and fix
Fewer log entries than expected	Retention is shorter than the requested window, or the level filter excluded them. Compare the requested window with the earliest returned timestamp before drawing conclusions
An error appears in the caller but not the dependency	Log level, handled errors, or retention — not proof the dependency is healthy. Widen the window or lower the level before concluding anything
Percentiles look implausible	Low request counts make percentiles unstable. Read them with the request count
Memory looks like a leak	A sawtooth is normal collection. A rising post-GC baseline is the signal; use get_memory_timeseries over a longer window
Metrics are empty but the app is running	Monitoring scope missing, or the window predates the deployment

Deployments

Symptom	Cause and fix
Nothing happened when I deployed	The call was a dry run. Mutating tools require confirm: true; the CLI requires the typed confirmation or --force. See Safety model
An infrastructure change was rejected	Redeploys of an existing app change the artifact only, by design. Use the scale or settings tools, or create a new deployment
delete_app fails with a deployment-ID mismatch	The deployment changed between your dry run and your confirmation. Re-run the dry run and use the new ID; this is the guard working
A production deploy refuses to proceed	Production requires an explicit acknowledgement. Provide it deliberately, or deploy to a lower environment first
Deploy succeeded but the app is unhealthy	A confirmed deploy of a broken artifact is still a successful deploy. Check get_app_status, then analyze_errors

MCP

Symptom	Cause and fix
The server does not appear in the host	Wrong file or wrapping key for that host, or the host was not restarted. See MCP setup
First call seems to hang	npx is downloading the package on a cold start. Point at a global anc install to avoid it
Every tool fails with an auth error	The server has no session. Run anc auth login in a terminal; the agent cannot and should not do it for you
The agent used the wrong environment	Environment is a parameter, not a default. Name it in the request
Tools resolve the wrong profile	The working directory is not the project you think. Check get_project_profile