My AI said it was not allowed to do that
What happened?
AI apps decide for themselves which actions need your say-so. Adding an inbox, reporting spam, or changing a setting can be refused when the app is not sure the request came from you, for example when it arrived through another chat, a schedule, or a message relayed from somewhere else. Inbox Triage never saw the request, so nothing changed on your account.
What do I do?
- Ask again yourself, in the same chat. A plain request in your own words ("add you@yahoo.com as an inbox") is what the app is looking for.
- Approve the prompt. Some apps show an Allow button when a connector wants to make a change. Choose it once for the action, or always for this connector.
- Do it on the website. Inboxes to add, connect, pause or remove an inbox; Settings to remove a VIP or edit your standing rules. Adding a VIP is done by asking your AI.
Claude Code and other command-line agents
These tools keep a permission list. To let this connector make changes without a prompt each time, add its tools to the allow list in the project's settings. The rule name depends on how you added Inbox Triage:
mcp__inbox-triage__add_inboxif you added it withclaude mcp add --transport http inbox-triage https://getinboxtriage.com/mcp.mcp__claude_ai_Inbox_Triage__add_inboxif it comes from your claude.ai connectors.
A rule with only the server part, such as mcp__inbox-triage, allows every tool from the connector. You
can also run the agent in a mode that asks you instead of deciding on its own. Your AI can tell you the exact rule
name for the action it was refused.
A link your AI gave you has expired
Connect links last 10 minutes. A password link also closes after 5 wrong tries. Ask for a new one, and for iCloud make sure you use an app-specific password. You can also choose Connect on Inboxes.
Try it with your own inboxes. Free for 7 days with up to 3 inboxes. No card needed.
Start free trialStill stuck? Email help@getinboxtriage.com.