Troubleshooting

Start with the symptom below. Your agent reads the evidence back for you through MCP or the CLI, and the links it gives you open read-only proof pages. Run anectico doctor first: it checks the API, sign-in, project, ingestion and MCP connection in one command.

No telemetry appears

Check the API key, project, endpoint, environment, time range, and SDK shutdown behavior. Ask your agent to run the instrumentation doctor (anectico doctor instrumentation).

No data is appearing

Telemetry appears but not under the customer

Check the stable ID, request scope, logout reset, and trusted propagation boundary.

Customer activity is not connecting

Replay is empty or stack frames are unreadable

Check browser capture or native replay ownership, consent and policy, release and distribution alignment, source maps, native symbols, and the recording's related evidence.

Replay or symbols are not working

Mobile flags or surveys fail

Check the explicit application endpoint and project, the separate client delivery keys, current evaluation identity and product consent. Keep an evaluated false, an unavailable decision, a valid empty offer list and a failed request distinct. The application renders surveys; receiving an offer does not display one. Read mobile flags and headless surveys and the platform guide before checking stored exposures and answers.

A plan or spend limit refuses work

Read the returned refusal and the current checked allowance. Follow plan headroom and agent budgets and typed spend for the available remedies. Unknown headroom is not zero, and an operation budget cannot raise a plan allowance. Review billing before changing a plan.

Alerts are not reaching someone

1. Test the notification destination directly.
2. Confirm the rule is enabled and uses the expected project and environment.
3. Check the threshold, evaluation window, and sustained duration.
4. Confirm the alert entered a firing state.
5. Verify the on-call contact method and escalation policy when used.

Alerts or integrations are not delivering

Sign-in, API, CLI, or MCP access fails

Use the status or client error to distinguish an invalid credential, missing scope, wrong project, or client configuration problem.

• Authentication or access is failing
• MCP or CLI is not working

Capture useful evidence

When contacting support, include:

• organization and project names, but no credentials;
• environment and exact time window with timezone;
• the proof-page link or the ID of the affected object;
• SDK language and version;
• the complete error code or message; and
• the smallest reproduction that still fails.

Never include an API key, session token, customer secret, or unredacted sensitive payload.

If the issue persists, follow Contact support to share that evidence safely. A registry 404 during early access does not mean the SDK is unsupported: use the artifact or repository access supplied during onboarding, as described in Choose an integration.