The Concierge model¶
Concierge is the escalation path from TalkIDE's AI team to a human engineer. This page explains what Concierge is, when it applies, and what its current availability looks like.
Availability
Concierge is not yet generally available to invited users. It is planned, and the concept described here reflects the intended design. The PM will confirm the real rollout status before this page is finalized. Treat the details below as how the feature is designed to work, not as a live promise.
What Concierge is¶
The AI team handles the vast majority of building and operating a project. But some situations require a human: a production incident with an unusual root cause, a compliance question, a complex infrastructure migration, or any case where the agents are blocked and cannot resolve the issue on their own.
Concierge is the mechanism for routing those situations to a human engineer on the TalkIDE team. You do not need to open a support ticket, find a contact page, or describe the problem from scratch. Mara escalates with full context: the project state, the error details, and a summary of what the agents already tried.
When escalation makes sense¶
The agents are designed to handle most situations autonomously. Escalation is appropriate when:
- A production issue has been ongoing and the automated recovery has not resolved it
- You need a guarantee about data handling or security that the AI team cannot provide
- Something in the infrastructure requires manual access that agents do not have
- You are in a regulated context and need a human sign-off on a decision
Mara will suggest escalation proactively if she detects a situation that fits one of these patterns. You can also request it directly by telling Mara "escalate to Concierge".
What Concierge does not do¶
Concierge is an escalation path, not a general support channel. It is not:
- A way to get a faster response than the AI team for routine questions
- A substitute for reading the docs on standard how-to tasks
- A channel for feature requests (those go through normal product feedback)
No SLAs or response time promises¶
During private preview, there are no published response time guarantees for Concierge. We do not have a number of engineers on standby that we can honestly quote. When Concierge becomes available, the terms and any response windows will be documented here.
How the concept fits the rest of the platform¶
The AI team page describes how the agents coordinate and where their boundaries are. Concierge is the designed exit point when those boundaries are reached. The goal is that escalation is rare, because the agents handle most things well, but the path exists and is never more than one Mara message away when it is needed.