If you’ve run a support operation on osTicket for any length of time, you know the pattern: the same questions come in week after week, agents spend valuable time composing replies from scratch, and consistency across the team is hard to maintain. After years of building and actively running osTicket-based support systems, I decided to solve this properly, with AI.
The result is the AI Reply Assistant Plugin for osTicket, a free, open-source plugin that uses OpenAI to generate draft reply suggestions for incoming support tickets, grounded in your own knowledge base.
How It Works
When a customer opens a new ticket or sends a reply, the plugin automatically generates a draft response and posts it as an Internal Note, visible only to your support staff, never sent to the customer automatically. Agents review the suggestion, adjust if needed, and send. That’s it.
For existing tickets, there’s also a manual ⚡ AI Draft button in the ticket action bar, so agents can request a fresh suggestion at any time without waiting for a customer message.
The AI doesn’t just generate generic responses, it pulls context from your knowledge base, canned responses, and osTicket’s built-in KB search, so suggestions are grounded in the actual information your team uses every day.
Key Features
- Automatic draft generation on new tickets and customer replies
- Manual ⚡ AI Draft button for on-demand suggestions in the ticket action bar
- Knowledge base integration, static mini-KB, osTicket KB search, and Canned Responses
- Smart filtering by department, priority, tag, and ticket status, so AI only activates where it makes sense
- PII redaction before any content is sent to OpenAI, emails, phone numbers, credit cards, personal IDs, and passwords are automatically stripped
- Configurable OpenAI parameters, model selection, max tokens, temperature
- Per-ticket and global rate limiting to control API usage and cost
- Audit logging without PII exposure
- Built-in admin UI for full configuration inside the osTicket Admin Panel
Why Drafts as Internal Notes?
This is a deliberate design decision. Fully automated AI replies in a support context are risky, AI can misread intent, miss nuance, or sound off-brand. By surfacing suggestions as internal notes, agents stay in control. The AI handles the heavy lifting of the first draft; the human handles judgment and the final send. In practice, this turns out to be the sweet spot: agents work significantly faster without the system ever sending anything a human hasn’t approved.
Requirements
- osTicket 1.18.x
- PHP 8.1+
- MySQL 5.7+ or MariaDB 10.3+
- PHP extensions:
curl,json,mbstring - An OpenAI API key
Note for Nginx users: you’ll need a small custom configuration to allow AJAX access to the draft button endpoint. Details are in the repository README.
Installation
- Copy the plugin folder to
include/plugins/in your osTicket installation - Go to Admin Panel → Manage → Plugins → Add New Plugin
- Install the plugin and create a new instance
- Enter your OpenAI API key and configure filtering preferences
- Enable the instance, done
Get the Plugin
The plugin is open source under GPL-2.0. You can grab it, fork it, and adapt it to your own osTicket setup.
👉 github.com/sasabajic/AI-Reply-Assistant-Plugin-for-osTicket
If you’re running osTicket and have been looking for a practical way to bring AI into your support workflow without losing control over what gets sent to customers, this is built exactly for that use case. Feedback, issues, and contributions are welcome.