Workflows
Playbooks, not pipelines.
Describe a process in plain markdown, or write it in JavaScript when it must run the same way every time. Typed inputs, schedules and event triggers, decision points that wait for a human, and artifacts you can trace.


How it works
Write the process. The agent runs it.
A workflow is a playbook prompt bound to an agent: the steps, the rules, the tone, in markdown anyone can read and edit. No node graphs, no YAML. Personal workflows for your own routines; shared workflows for the whole workspace.
When a process is exact, write a Code workflow instead: JavaScript that calls the same tools, uses the model as one deliberate step, and can wait for a timer, an answer, an approval, or a coding run, even a week later. A playbook can be converted to Code in place, keeping its triggers and run history.
Flagship example
AI delivery templates: from task to reviewed pull request
The AI delivery templates are short Code workflows that carry an agent's work through review: task triage and start, request refinement, AI review, fixing AI findings, CI repair, reviewers and reminders, handing a review to the agent, risk and merge, and after merge. Each reacts to one event and finishes; when the next step depends on CI, a review, or a push, the workflow for that event takes over.
Nothing is installed in a project by itself. Create the templates a project needs, one at a time: each says which others it works with, and Set up triggers wires its events to that project in one step. Every workflow you create is your own copy to change.
An agent's approval never counts as a person's, and the branch protection on your code host stays the floor. See how the loop fits together on AI coding.
Capabilities
Built for real processes
Triggers
Run manually, on a schedule, or from events: tasks, requests, pull requests, CI and AI reviews, support tickets, channel messages, and alerts.
Start on an item
Start a workflow by hand from a task, request, pull request, file, or support ticket, with that item filled in. A reply to a ticket’s customer waits on a card only you can send.
Typed inputs
Structured parameters, including references to Volobox items.
Human decision points
The agent pauses with a structured question; the decision lands in your Inbox.
Journey & artifacts
Every run has a traceable journey, transcript, and produced artifacts.
Coding handoff
A workflow can open a coding session and carry work all the way to a PR.
Alerts
A run can raise an alert for a condition that lasts, such as a nightly check that keeps failing, and resolve it once a later run finds it fixed. A workflow can also start when one of your insight alert rules fires.
Learn moreRuns in the Story
A run started by an event on a task, request, pull request, or ticket shows in that item’s Story, with what it created.
Dry Run and Rehearse
Dry Run checks the reads. Rehearse records every write the workflow would make, without making it.
At a glance and Storyboard
Read-only views of what a workflow starts on, reads, writes, and stops on, derived from the body as you type.
Fill with AI
Describe the process and the assistant drafts the playbook or the code, checked before it reaches the editor.
Cost per run
Every run shows its runtime and models, and its AI cost to the people allowed to see it. Administrators can cap the spend of a single run.
Example
Release QA test-scenario generation
Point the workflow at a release; the agent reads the changes, drafts test scenarios as structured artifacts, and pauses for review where it matters. One playbook, repeatable every release.
Workflows run under the agent's autonomous authorization profile: approvals, limits, and audit included.
Where it fits
Automate the work between the systems.
CI/CD moves the code, and your tracker keeps the record. In between sits the human work: reading what changed, drafting the plan, asking for a decision, writing it down, telling the team. Playbooks automate that layer. It does not replace your pipeline; it runs the process around it.