Skip to main content

Support · Agents

How to Create an Agent

Configure and deploy an autonomous AI agent that works on your behalf — from simple automations to complex multi-step workflows.

Back to Support

What is an agent?

An agent is a persistent AI that runs tasks on your behalf, either on a schedule or in response to a trigger. Unlike a single chat message, an agent can execute multi-step workflows, check external systems, make decisions, and report back — all without you needing to be present.

Agents operate inside the TaurusX Execution Engine (TEE), which enforces your rules and the platform's safety policies at every step.

Agents are available on Pro, Pro+, Enterprise, and Enterprise+ plans. Free plan users can chat with TaurusX directly but cannot deploy persistent agents.

Create your first agent

To create a new agent:

  1. Open the Agents tab from the main navigation.
  2. Tap New Agent in the top right.
  3. Give the agent a name that describes what it does (e.g. "Daily Standup Summariser").
  4. Write a responsibility — one or two sentences explaining what this agent's job is.
  5. Write instructions — the full prompt that tells the agent how to behave, what inputs to expect, and what outputs to produce.
  6. Choose a trigger: Manual, Schedule, or Event-based.
Be specific in the instructions. Instead of "summarise Slack messages", write "Every weekday morning, pull the last 24 hours of messages from the #engineering channel and write a 3-bullet summary of blockers, decisions, and action items."

Set permissions

Before deploying, select which tools and integrations the agent is allowed to use. TaurusX shows a checklist of all connected services. Only enable what the agent genuinely needs.

Permission levels per integration:

  • Read-only — agent can fetch data but cannot write or modify
  • Read + Write — agent can create, update, and delete records
  • None — integration is not accessible to this agent
GOVMODE will block any action that exceeds the permissions you set here, even if the agent's instructions ask for it. Keep permissions as narrow as possible.

Deploy and run

Once configured, tap Deploy Agent. The agent enters a review queue — on accounts with GOVMODE enabled, an approval may be required before the agent becomes active.

Once approved the agent status changes to Active. You can trigger a manual run at any time from the agent detail screen, or wait for its scheduled trigger to fire.

Each run produces a detailed execution log: inputs received, steps taken, outputs produced, and any errors encountered. Logs are retained for 30 days.

Monitor and pause

The Agents tab shows a live health badge for every deployed agent: Healthy, Degraded, or Critical. Tap any agent to open its detail view with:

  • Real-time status and last run timestamp
  • Run history and execution logs
  • Health score (0–100) based on recent performance
  • Risk level assigned by the platform

To pause an agent without deleting it, tap Pause from the detail view. The agent retains its configuration and can be resumed at any time.

Still need help?

Our support team usually responds within a few hours.

Submit a ticket