For agents
Automate any business process. With business context.
Canopy Agent Workflows chain triggers, agents, approvals and actions on top of Canopy, so every automated step knows the metric, the driver, the owner and the plan. The metrics start them, humans gate them, agents run them end to end.
Automation with context. Not another cron job.
The metrics start the process. Business-aware triggers
Missed targets, thresholds, outliers and silent metrics fire workflows, alongside schedules and inbound webhooks when you want them.
Humans gate what matters. Approvals first-class
Named reviewers, explicit rejection paths, and escalation up the real org chart when nobody responds.
Fully agentic where you choose. Your agents as steps
Drop platform or custom agents into any flow and fan out across your systems with the HTTP action.
Triggers that watch the business, not the clock
Generic automation fires on schedules and webhooks; these workflows also fire on the business itself. A target missed at period end, a threshold crossed, an outlier detected, a metric gone silent, a RACI assignment made, a task created or overdue. Schedules and inbound webhooks are there when you want them, but the point is that the metrics themselves can start the process.
Build it visually. Publish it live.
The full palette, exactly as it appears on the canvas. Nine triggers watch metrics, tasks, ownership, schedules and webhooks; steps run agents, query metrics, branch, delay and wait for approval; actions post to Slack, send email, create tasks, assign RACI, escalate ownership and call any external system. Chain them on the canvas and publish: live enrolments re-flow, and every run records the exact path taken.
Triggers
Steps
Actions
Re-score stalled opportunities
Reforecast deferred revenue
Brief account owners
Start from a template
The gallery ships production-shaped starting points: the Monday action plan to Slack, the monthly RACI audit email, stale-data and missed-target escalation up the org chart, threshold-triggered RACI reassignment, webhook-to-partner-API fan-out, and an approval policy that walks up the reporting line on its own. Pick one, adjust the nodes, publish.
Weekly action plan to Slack
Runs the action-plan agent every Monday and posts a summary to a Slack channel.
Monthly RACI audit email
First of each month, audits RACI assignments and emails a digest of the gaps.
Stale data: owner, then manager
Pings the metric owner when their data goes stale; if nothing changes, escalates to their manager via the org chart.
Target missed: escalate up the chain
Fires at period close when a metric misses its target. Notifies the Responsible owner, then their manager.
Threshold breach: reassign Accountable
When a metric crosses a threshold, moves the Accountable RACI to the manager of whoever currently owns it.
Approval with escalation
An approval that walks up the org chart on its own: notify the manager, reassign, then auto-decline if nobody acts.
Inbound webhook to partner API
Forwards an inbound webhook to a partner system over authenticated HTTP, with the body templated from the payload.
One-off investigation
Manually triggered agent run with a Slack summary. Useful for ad-hoc deep dives.
Rich business context at every step
Steps read Canopy as they run: the driver behind the move, the Accountable owner, the verified outcome history, the plan. Workflows understand your org chart natively, so you can escalate to a user's manager, assign RACI by job title, or find the common manager of a team without hardcoding a single name.
Channel
#revenue
Body
{{ trigger.metric.name }} missed target: {{ trigger.metric.value }} vs {{ trigger.metric.target }}
Plan: {{ steps['action-plan'].output_markdown }}
Owner: {{ metric.accountable.name }}
Live values from the trigger, upstream steps and Canopy context.
Humans stay in the loop, by design
Approvals are first-class: the next action only runs if the named reviewer approves, rejections branch to their own explicit path, and approvals can escalate up the org chart automatically when nobody responds, with the policy visible on the run itself. You decide which steps need a person and which do not.
Roll out the revenue action plan?
Stress-tested by the action-plan agent. Conversion rate is the driver; three tasks proposed.
Waiting on Sarah ChenIf nobody responds, this notifies Sarah's manager, then reassigns, then auto-declines.
Then let agents run it end to end
The same plan a person would approve can fan out across your systems: re-score stalled opportunities in your CRM, reforecast deferred revenue, brief the account owners in Slack. Deploy your own agents and drop them in as steps alongside the platform ones, so the whole process can be fully agentic where you choose. And the HTTP action calls any external system with saved auth profiles and tunable retries, so the workflow's reach is your stack, not our integration list.
Re-score stalled opportunities
Reforecast deferred revenue
Brief account owners
Every action traced back to the run.
Every run tells you exactly what happened
Workflows keep their own run history: every fire shows the path taken through the graph, the inputs at each step and the outputs produced. When a run is waiting on someone, the run page says who and what happens next. No piecing it together from logs.
Wait for Sarah's approvalUtilities