Automations

Key Concepts

An automation couples a trigger to a step graph. When a matching event fires, the automation starts a run that walks its graph, evaluating conditions and executing actions along the way. Automations are the platform's answer to "when X happens, do Y" without standing up a bespoke service.

Step Graphs

A step graph is a directed graph of steps. Each step is a condition (branch on data), an action (invoke an agent, send a message, transition a workflow item, write a row), or a wait (pause until a resume key arrives). Edges connect steps into the control flow the run follows.

• Conditions gate execution on the triggering payload or accumulated run state.
• Actions call other services as the automation's principal, using cross-service grants.
• Waits suspend the run until an external event (a resume key) wakes it.

Triggers

An automation subscribes to one or more event types emitted by any service — an inbound message on Channels, a workflow item transition, a new row on Worksheets, and so on. Time can also be a trigger: schedule an automation to run on a cadence. Depth and trigger budgets keep reaction chains from running away.

# List automations for a project
curl "https://platform.ergondata.ai/api/v1/automations/projects/{project_id}/automations" \
  -H "Authorization: Bearer {token}"

Runs & Resume Keys

Each firing produces a run — an auditable record of which steps executed, with what inputs, and what they returned. When a step waits, it registers a resume key; delivering that key later resumes the run exactly where it paused. Resume keys are how long-running, multi-actor choreographies stay durable without polling.

Connections

Actions reach other services through the platform's unified access model: the automation is a principal, and each target resource is shared with it via a grant or a connection request. This keeps every automated side effect scoped and auditable, exactly like a human or agent performing the same action.