Take 10 minutes.
Get hours back every week.
Short walkthroughs that turn the platform into your operations team: automated pipelines, instant follow-ups, AI-handled admin. One video, one less thing to do manually.
Understanding and editing your scheduled agent
How to read the configuration sidebar on a scheduled agent's detail page, trigger a manual test run, and use the settings editor to update its kickoff prompt, schedule, persona, tools, and model.
Clicking a scheduled agent in Agent Hub opens its detail page at https://app.trustpager.com/auto/agent-hub/scheduled/{agent_id}. The right-hand column is a Configuration sidebar that shows everything about how the agent is built and wired — no digging through settings menus required. From the same page you can fire a one-off test run or jump straight into the settings editor.
The Configuration sidebar
The sidebar is divided into twelve accordion sections. You can expand any section independently.
Information
The top section shows the basics at a glance: the agent's display name, type (scheduled), runtime, trigger kind, current status, and its unique agent ID. The agent ID is useful if you ever need to reference this agent in API calls or automations.
Schedule
Shows the cron expression, a human-readable summary of when the agent runs (for example, "Every day at 6:00 AM"), the timezone it runs in, and timestamps for its last run and last successful run. The Edit schedule button opens a modal where you can change the timing without leaving the detail page — though if you want to change the timezone at the same time, use the full settings editor instead (see below).
Authorization
Lists the OAuth connections the agent is running under and the scopes it has been granted. If a run fails with a permissions error, this is the first place to check — the agent may need an additional scope or its OAuth token may need to be reconnected.
Model
Shows which AI model the agent uses (for example, Claude Opus 4.7 or Claude Sonnet 4.6), its speed setting, and the current version number. You can change the model in the settings editor.
Persona
Displays the first 200 characters of the system prompt that shapes the agent's behaviour and tone. Click View full to read the entire persona in a modal, with a copy-to-clipboard button for easy reference.
Kickoff prompt
A preview of the message the agent receives at the start of every run — this is where you embed runtime context like today's date or the company name. Click Edit kickoff to open the kickoff editor inline. Changes here are saved as a draft and do not bump the agent's version number.
MCP Servers
Lists the integrations the agent connects to on each run. These are the external systems (your CRM workspace, calendar, email, and so on) the agent is authorised to read from and write to.
Tools
Shows every tool the agent can call, with a permission policy chip next to each one. A green always allow chip means the agent calls that tool automatically. An amber always ask chip means the agent will pause and wait for your approval before using that tool. You can review and change tool permissions in the settings editor.
Skills
Lists any skill catalog entries attached to the agent — reusable instruction sets that give the agent a specific capability without having to re-write it into the persona each time.
Vaults
Shows credential bundles the agent has access to. Vaults let you give the agent access to sensitive credentials (API keys, passwords) without embedding them directly in its prompt.
Capabilities and wiring
Shows the capability tags registered for this agent and which other agents in your workspace can signal it. This section is most relevant for multi-agent workflows where one agent hands off work to another.
Configuration and Developer details
The bottom two sections expose the raw JSON configuration for the agent. These are aimed at power users who want to inspect or copy the exact spec — for example, when building a sibling agent with a similar setup.
Running a manual test
The Test Run button in the page header fires the agent immediately, exactly once, outside its normal schedule. Use this after you've made changes and want to verify the agent behaves correctly before the next scheduled run arrives.
You need permission to manage agents to use Test Run. If you don't see the button, ask your workspace admin to grant you agent management access.
Good habit: After editing the kickoff prompt or schedule in the settings editor, click Test Run to confirm the agent picks up your changes and runs without errors. Check its output in https://app.trustpager.com/auto/agent-reports a minute later.
The settings editor
Click Open Settings in the page header to open the full settings editor at https://app.trustpager.com/auto/agent-hub/scheduled/{agent_id}/settings. The editor shows the agent name, a "Last updated" timestamp, an unsaved-changes indicator, and the current version number (for example, v3) across the top toolbar. There are five tabs along the top, each covering a different aspect of the agent's configuration.
Kickoff tab
A full textarea editor for the kickoff prompt the agent receives at the start of every run. Four variable chips sit above the textarea — click any of them to insert a placeholder:
- {{today}} — replaced with the current date at runtime
- {{company_name}} — replaced with your workspace name
- {{agent_registry_id}} — the agent's unique identifier
- {{trigger_context}} — contextual data passed in when the agent is triggered
Below the editor, a live preview shows exactly what the agent will see after variable substitution — so you can confirm the prompt reads correctly before saving.
Click Save Draft to save your changes. No version bump — the next scheduled run uses the updated kickoff prompt straight away.
Schedule tab
A visual cron builder where you set how often the agent runs, with a timezone dropdown alongside it. A "next 10 fires" preview updates as you adjust the settings, so you can see exactly when the agent will next run before committing.
Click Save Draft to save your changes. No version bump.
Persona tab
A full textarea editor for the system prompt that defines the agent's role, tone, and behaviour. The current version badge sits above the editor as a reminder that saving here has a bigger effect than saving the kickoff or schedule.
Click Publish to save your changes. Publishing creates a new version of the agent. Any run that was already in progress continues on the previous version; only runs that start after you click Publish use the new version.
Tools tab
JSON editors for the tools array and MCP servers array, with live validation so you can see any formatting errors before saving. Use this tab when you need to add, remove, or reconfigure a tool or integration at the spec level.
Click Publish to save your changes. A new version is created.
Model tab
A model dropdown (Claude Opus 4.7, Claude Sonnet 4.6, or Claude Haiku 4.5) and a speed dropdown (default or fast). Changing the model affects every run from the next Publish onward.
Click Publish to save your changes. A new version is created.
Save Draft vs Publish — what's the difference?
The two save actions have different weights:
- Save Draft — persists lightweight changes (kickoff prompt, schedule, timezone). No new version is created. The next scheduled run picks up the change automatically.
- Publish — persists heavyweight changes (persona, model, tools, MCP servers) and creates a new version. Any run already in progress finishes on the previous version; only new runs use the published version.
You need permission to manage agents to use Save Draft or Publish. If you don't see these buttons, ask your workspace admin to grant you agent management access.
Related pages
- Agent Hub list view — see all scheduled agents, their run status, and navigate to any detail page. Open at https://app.trustpager.com/auto/agent-hub. Full walkthrough: How to Monitor AI Agents in Agent Hub.
- Agent Reports — read the output from each run after it completes. Open at https://app.trustpager.com/auto/agent-reports.
- Agent Alerts — configure notifications for failures, missed runs, and error spikes. Open at https://app.trustpager.com/auto/agent-alerts.
Two ways to
meet FinalPiece.
Talk to Evie right now, or book a real human for a deeper walkthrough. Whichever feels right.
Pick up the phone.
Hear her in 10 seconds.
- RealHer voice passes for human.
- SmartKnows your prices, hours, and trade.
- PatientPolite, attentive, never has a bad day.
Pencil us in.
60 minutes with a real human · Google Meet · Free.