Triggers, schedules, and multi-agent workflows
So far, we've covered running agents manually - either from within Cortex or by dispatching them from Street. But Cortex also lets you automate when agents run, using triggers and schedules, and even have agents trigger other agents.
Triggers: running agents automatically on Street events
Instead of dispatching an agent manually every time something happens, you can set up a trigger so that the agent runs automatically whenever a specific event occurs in Street.
To add a trigger, open your agent's settings and navigate to Triggers. You'll see the full list of Street events that are available. For example:
Enquiry booked - The agent runs every time a new portal enquiry comes in.
Applicant registered - The agent runs whenever a new applicant is added.
Offer received - The agent runs when a new offer is logged.
Once you save a trigger, the agent will fire up and run its process automatically each time that event occurs - no manual intervention required.
This is ideal for workflows that should happen consistently every time. For example, a portal enquiry management agent with an "Enquiry booked" trigger will handle every new enquiry the moment it arrives, around the clock.
Schedules: running agents on a timetable
Not every workflow is event-driven. Some tasks need to happen on a regular cadence - weekly, monthly, or at a specific time. For these, you can schedule an agent to run automatically.
When adding a trigger, instead of selecting a Street event, choose Schedule. You can configure:
Frequency - How often the agent runs (e.g., daily, weekly, monthly).
Day - Which day of the week or month.
Time - What time the agent should start.
For example, you could schedule a local area content agent to run on the 20th of every month at 9am, so you always have fresh blog articles without anyone needing to remember to trigger it.
Agents triggering other agents
One of the most powerful features in Cortex is the ability for one agent to trigger another. Using the Trigger agent tool, you can create multi-agent workflows where a coordinator agent manages the overall process and delegates specific tasks to specialised agents.
How it works
Give your coordinator agent the Trigger agent tool.
On the agent you want to be triggered, go to Settings and specify which agents are allowed to trigger it.
When the coordinator runs, it can trigger one or more agents as part of its workflow.
Example: local area content
Imagine you cover four local areas and want a blog article written for each one every month. Rather than creating four separate scheduled agents, you could:
Create a Local Area Content Agent that writes a blog post for a given area.
Create a Local Area Content Coordinator that triggers the content agent once for each area you cover.
Schedule the coordinator to run monthly.
Each month, the coordinator runs once and triggers four separate content agents, each writing an article for a different area.
Other multi-agent patterns
The coordinator pattern is just one example. You can also build:
Classifier agents that assess incoming data and trigger different agents depending on the outcome (e.g., routing enquiries to different agents based on enquiry type).
Sequential pipelines where one agent completes a task and then triggers the next agent in the chain.
Fan-out workflows where a single agent triggers many parallel agents to process a batch of work.
The flexibility to chain agents together means you can build workflows of almost any complexity, while keeping each individual agent focused and manageable.
Street is constantly being updated. This can mean that some of these articles become slightly out of date. If you spot something that looks a little bit different to your Street account, please let us know on the live chat so we can get it updated!