Building a Spell Check agent from a plain-English description
Agents are currently in Beta. Each run consumes credits from your workspace balance, and the exact cost is shown before you confirm a run.
The four ways to work with agents
Build one from a description
Write the check in your own words. Superflow stress-tests it against sample cases and shows you where it would succeed or produce a false positive, before it ever runs on a real project.
Import a checklist you already have
Upload a PDF, CSV, Excel or text checklist. Superflow reads every task, groups the related ones, and proposes a set of agents that cover them.
Group them and run them together
Bundle the checks you always run, such as a pre-launch set, and trigger the whole group at once.
Run one against a project
Pick a project and a path, choose desktop or mobile, and review the findings. Every past run is kept so you can compare.
Reviews that start before you have finished installing
Add a URL when you create a project and the agents you select start reviewing the site immediately, while you are still copying the script onto it. The site only has to be reachable.Agents reviewing a site during project setup, before the toolbar is installed
Run agent pre-checks on a new project
The full walkthrough, step by step.
Turning an existing checklist into agents
If your team already has a QA checklist, you do not have to rewrite it. Superflow reads the file, works out which tasks can be verified on a rendered page, and tells you which ones cannot. Settings inside a CMS, for example, are flagged as excluded rather than silently dropped.A QA checklist being imported and turned into seven agents
Running one against a project
Pick the project and the path, choose desktop or mobile, and Superflow shows the exact cost against your balance before you confirm. Findings come back as the issue, a suggested fix, and the markup it sits in. Every run is kept, so you can compare a new one against what the agent found last time.Running an agent against a project and reviewing its findings
How to Run an Agent from the Agents Tab
The full walkthrough, step by step.
Running a whole set at once
Creating an agent group and running every agent in it together
Running them from your issue tracker
Everything above is a run you start yourself. An agent workflow starts the run for you: a ticket moving into QA dispatches the agents, the findings land back on that ticket as a comment, and the card moves to your fix column — or to Done when every agent comes back clean.A ticket entering QA dispatches the agents; the verdict moves the card
Agent Workflows
Build the loop on Jira, or drive the same engine from your own system.
Sites behind a login
An agent can only report on a page it can actually read. If your site sits behind a password, a login, or single sign on, configure Site Access in your project settings first — otherwise the run stops and tells you, rather than describing the sign in screen as though it were your site.Letting Superflow read your protected site
Work out which of the five options you need, then set it up.
AI in the toolbar
Agents review pages. The AI Co-Pilot works the other way round: highlight copy on a live page, get rewrites, and turn the one you like into a task.How to Use the AI Co-Pilot
Highlight more than two words to bring up the AI toolbar.