Platform6 min read

Managing Agents

Monitor, configure, and manage your agents from the Symbi dashboard.

The Agents Page

The Agents page lists everyone your company has hired. Each row shows the agent's name, role, current status, and key performance numbers. Click any agent to open their detail page.

Status badges in the agent list tell you whether an agent is:

  • Active — listening for triggers and processing tasks
  • Paused — temporarily stopped; no tasks will be picked up

Dashboard Overview

The Dashboard gives you a unified view of all your agents. Four metric cards at the top summarize performance:

  • Interactions — total tasks processed across all agents
  • Hours Saved — estimated time saved compared to manual handling
  • Cost Saved — financial value of those saved hours based on your configured rates
  • Success Rate — percentage of actions approved without edits

Use the date range filters to focus on a specific period and see how metrics compare.

Agent Detail Page

Each agent has one page that scrolls through its sections. A side bar (or, on narrow screens, a row of links under the header) jumps straight to each one. The header holds the agent's main actions: Publish changes when there is something to publish, Pause / Resume, and Run now for scheduled and manual agents.

  • Apps — which of your company's apps the agent can use, and the specific resources (queues, labels, spreadsheets) it reads from and writes to
  • The Job — the agent's title and instruction, and the Runs when line that says what starts it. See Tuning Your Agent's Behavior
  • What it does — the agent's capabilities and the review mode for each action group
  • Approvals & knowledge — where the agent's actions are reviewed, and which knowledge bases it may search
  • How it communicates — tone, language, and other reply preferences
  • How it works and Who it is — the agent's operating manual and persona. Both start out written by Symbi, and are grouped apart so you can see which parts of the page you didn't write
  • Versions — which published version is running, and the way back to an earlier one

Everything above this point is part of the agent's draft and takes effect when you publish. Below a Settings — not part of any version divider are the parts that save immediately: this month's performance, the agent's API, its template details and performance estimate overrides, and the Danger Zone for deleting the agent.

Setting Up a New Agent

Until an agent is published for the first time, the side bar shows a three-step checklist in place of the section links:

  1. Give the agent apps — pick the apps and resources it works with in the Apps section, and choose under Runs when what starts it
  2. Describe the job — write a sentence or two about what the agent should do for your company, then let Symbi draft the job title, instruction, operating manual and persona from it. You can edit the result before you publish
  3. Check what the agent can do — read through the capabilities in What it does and confirm they look right

When the agent has a name and a job, press Publish & go live (v1). If something still blocks publishing — for example, an agent that runs on incoming events but has no app to bring it work — the button stays disabled and says what is missing. The agent starts working on real items as soon as publishing succeeds.

Versions and Publishing

Every set of changes you make to an agent — including which apps it connects to and which specific resources (queues, labels, spreadsheets) it uses — is held as a Draft until you publish it. To release updates to production, click Publish changes on the agent's detail page and enter a brief release note.

Publishing creates a new numbered version (v1, v2, …) and switches all incoming triggers to the new configuration immediately — the next inbound ticket or webhook runs against it, so there is no drain window. The dialog shows what changed before you confirm; updated app connections appear under Apps alongside any other changed settings. A lifecycle badge in the page header confirms the current state:

  • Draft — unpublished; no automatic triggers fire
  • Live (vN) — the published version running in production
  • Testing (vN) — published to a test slot; manual runs only
  • Paused — suspended; configuration preserved until you resume
  • Deprecated — retired; no new tasks will start

Info

Agents that existed before versioning was introduced start as Live and continue running without any action needed. Configuration edits move the agent to Draft until you publish again.

Running Tasks

Agents start work when a trigger fires:

  • Event-driven agents run automatically when something happens in an external service (e.g., a new support ticket arrives)
  • Scheduled agents run on a schedule you've configured (e.g., every morning at 9 AM)
  • Manual agents run when you click Run now in the header of the agent's page

For scheduled or manual agents, the Run now button lets you trigger a task on demand — useful for testing, one-off work, or verifying behavior before a scheduled run. Run now is blocked for paused agents — clicking it while the agent is paused shows an error. Resume the agent first, then trigger the run.

Activating and Pausing

You can toggle any agent between Active and Paused at any time.

  • Active agents listen for triggers and process tasks per their configuration
  • Paused agents stop entirely — no tasks will be picked up until you reactivate them

Tip

Pausing is useful when you want to update configuration or a knowledge base without disrupting live work. Reactivate once you're ready.

Configuring Your Agent

From the agent's page, company admins can change which apps and resources the agent uses, its job title and instruction, what starts it, its capabilities and review modes, the knowledge bases it may search, its communication preferences, and its operating manual and persona. Other members see the same page read-only. Knowledge bases themselves are managed on their own management page. Your Personal Assistant can help with this too — just open the chat panel and ask it to change the setting on your behalf.

What's Next?