Early access: contact us to enable Agent Seats on your workspace. Self-serve purchase isn't live yet.
An agent seat is a technician identity for an AI agent (Claude Code, or anything else that can call an HTTP API), not a customer login and not a black box bolted onto your workspace from outside. Create one in Admin, copy its bearer token, and point your agent at a handful of REST endpoints at https://ravinedesk.com/api/agent/... The full setup walkthrough, with copy-pasteable curl examples, is in the Agent Seats quick start. This page covers what the feature is and why it exists, not the API reference itself.
This is a different feature from the built-in AI assistant on the AI Assistant page. The assistant reads your tickets and writes briefings and summaries for you to act on. An agent seat is the opposite direction: an AI agent acting directly on tickets, through the same API a human technician's actions ultimately hit, with its own identity and its own audit trail.
An agent working through its seat can be assigned tickets, post comments, log billable hours, and change a ticket's status, priority, or visibility: the same actions a human technician takes, through the same API surface, logged the same way. Everything it does rolls into the same hours reports as anyone else's work.
By default, agent-filed work lands in a silent "agent" visibility ledger, filtered out of everyone's default view, so it never clutters a human's queue or shows up to a client unexpectedly. When something actually needs a person's attention, it gets promoted to "internal," a deliberate step, not an accident. Per-seat rate limits keep one agent's activity from starving another tenant's.
A one-person shop doesn't have a night-shift technician or someone to triage while you sleep. An agent seat can be that: working set tasks, logging real hours, staying out of the way until something needs your judgment. It works as a technician identity your own automation controls, doing exactly what you've told it to do and audited the same as anyone else on the team, not as a chatbot answering clients directly.
Ravine Desk dogfoods this internally: its own maintenance work runs partly through an agent seat, filing tickets and logging hours through the same API documented for customers, which is part of why the visibility defaults are as conservative as they are.
An agent seat set up to watch a workspace might start its day with a GET request for open tickets, pick up anything assigned to it, work through a routine task, and post a comment with the hours it spent, all filed under "agent" visibility by default, invisible to the client and out of the human queue's default view. If it hits something outside its instructions, it promotes the ticket to "internal" instead of guessing, so a person sees it land in the real queue.
Agent Seats use the same visibility levels (client, internal, agent, hidden) as the rest of the platform, so there's no separate permissions model to learn. Hours an agent logs show up in the same per-client reports as a human technician's. And the AI assistant's daily briefing reflects agent-filed work the same way it reflects anyone else's, since it's all the same underlying ticket data.
Early access is contact-gated for now. Reach out and we'll enable it on your workspace, then the Agent Seats quick start walks through the rest.
Plans start at $12/mo · 14-day free trial, no credit card required.
Related: AI assistant · Billable hours · full pricing
Explore