AI Conversations
AI Conversations provide an agent workspace inside Testable Monitoring. Use it to ask questions, investigate issues, summarize account health, and take action across the same monitoring resources you manage in the web app.
In Agent mode, AI Conversations are designed to support the same kinds of actions you can perform in the monitoring web app, including creating and updating monitors, reviewing checks and pings, managing incidents, configuring maintenance windows, working with status pages, changing notifications, and updating organization-level monitoring settings.
Open AI Conversations
Select AI Conversations from the monitoring app sidebar. The page opens a full chat workspace with a message composer, model selector, runner region selector, mode selector, file upload action, settings button, conversation history panel, and a copyable Monitoring MCP server URL. The same workspace can also open as a sidebar from other monitoring pages.
What You Can Ask
The agent understands monitoring concepts and can work across the full monitoring workspace. The examples below are starting points, not a complete list of supported actions.
- Summarize monitors - Ask for current status, trends, and areas of concern.
- Create monitors - Describe the target, scenario, checks, schedule, and region in plain language.
- Change monitor settings - Ask for bulk changes such as moving 5-minute monitors to 15-minute schedules.
- Manage notifications - Add teammates to monitor notifications or explain who will receive alerts.
- Create maintenance windows - Describe planned downtime and any monitors, folders, groups, or tags to include or exclude.
- Investigate incidents - Ask the agent to review recent incidents and suggest where to focus next.
- Manage status pages - Ask about public or private status page configuration, associations, visibility, and history settings.
- Update organization settings - Ask for help with shared defaults and monitoring configuration that would normally be managed in the web app.
Conversation Modes
- Agent - Can make changes to your monitors and related configuration when tool calls are approved.
- Ask - Read-only mode for questions, summaries, and explanations.
Models and Regions
The model selector can be left on automatic selection or set to a specific available model. The region selector lets you provide regional context for monitor creation and investigation, using the same monitoring regions available elsewhere in the app.
Tool Approval
When the agent needs to call a tool, Testable shows the tool call in the conversation. Approve individual calls when you want the action to run. For repeated trusted actions, use Always Approve; those tools appear in the settings modal under Auto-Approved Tools and can be removed later.
The agent can only act within the permissions, plan features, and rollout state available to your organization. Use the tool approval flow to review changes before they are applied, especially for create, update, pause, resume, and delete actions.
Conversation History
Conversations are saved in the right-side panel. You can create a new conversation, reopen previous conversations, rename conversations, and delete conversations. Conversation URLs include a conversation ID so a specific conversation can be reopened directly.
Attachments and Browser Sessions
Use the upload action to attach files to the active conversation. When an agent uses browser-based tools, the conversation header can show browser session activity and available artifacts. The conversation also tracks metadata such as elapsed time, model, cost, and context usage when available.
AI Settings
Open the settings button beside New Conversation to copy the MCP server URL, configure AI settings, and manage auto-approved tools. AI settings include using your own OpenAI key, optionally setting a custom OpenAI-compatible base URL, and defining an organization-level system message. The auto-approval section can approve every tool or remove individual tools previously added with Always Approve.
Availability
Starting a conversation requires a confirmed email address, available AI credits, and write access to the organization. Viewer accounts can read existing conversations but cannot send messages or approve tool calls.