Admin tab: Agents (configuration in 8 tabs)

Agent management (menu item Agents, administrators only) leads to a detail page per agent, organized into eight tabs. This page is the configuration reference; the concept behind it is explained in Agent model, and the quick start in Create an agent and test it in a dry run.

The eight tabs

1. Overview – name, description, kind (identifier, e.g. lead-screener), model (provider + model name), max iterations (1–50), max runtime in minutes (1–240), Enabled.
2. Scope – source adapter (depending on the registered adapters) and the associated scope editor.
3. Reasoning – the agent’s system prompt (Markdown, up to 2000 characters).
4. Skills – skills via multi-select, grouped by category.
5. Tools – tool selection with intent labeling (READ/WRITE); the Working folder root field is desktop-specific.
6. Triggers – one or more triggers per mode: Schedule (cron), ChatMessage (conversation selection), Manual and ManualDryRun (prompt); per trigger Run / Dry run / Save / Delete as well as “last/next run”.
7. Identity – Run-as user (execution identity) and Owner user.
8. Runs – run history (ID, mode, status, start, duration, tools) with pagination.

Limits & notes

• Requires the Administrator role.
• The Working folder (Tools tab) only takes effect in desktop mode.
• Scheduled agents in desktop mode only run while the app is open; for continuous operation use web mode. → Web vs. desktop
• Identity & permissions follow the role/confidentiality model. → RBAC & confidentiality levels

Related

• Agents (feature) · Agent model
• Create an agent and test it in a dry run
• Set up a cron trigger
• Enabling agent delegation from chat