Getting started

Workflowware is usable even if you know nothing about AI agents. You do not need MAAT Runtime, agent frameworks, model providers, prompts, or code to test a package.

Also available as markdown: customer-getting-started.md

What am I buying or downloading?

You are getting a workflow package: a folder of instructions, examples, rules, and test cases that explain how a specific job should be performed by humans and AI agents together.

missed-lead-recovery-workflowware/
  README.md
  workflow-spec.md
  agent-spec.md
  approval-rules.md
  eval-template.csv
  agent-handoff.md
  artifact-manifest.json

A package is not an app-store app. It is closer to a complete operating manual plus test kit for one job.

Where do I download it?

Early public downloads may come from the Workflowware website, the propershare/workflowware GitHub repository, a GitHub Release ZIP, or a direct link during a pilot.

If you are not technical, choose the guided pilot instead of installing files yourself.

Can I use this on a phone?

Yes, for review and guided testing. On a phone you can usually read the package page, fill intake, submit sample data, review drafts, approve or reject a message, and give feedback.

A phone is usually not the best place to unpack ZIPs, edit package files, run scripts, or connect business tools.

TaskPhoneComputerGuided pilot
Read the websiteYesYesYes
Fill out intakeYesYesYes
Submit 3–5 sample leadsYesYesYes
Review outputsYesYesYes
Download ZIP packageMaybeYesNot required
Edit package filesNoYesNot required
Connect business systemsNot recommendedMaybeBest

What if I know zero about agents?

Use the guided pilot path. Answer plain business questions:

  • What job do you want help with?
  • What does a good result look like?
  • What should require human approval?
  • What should never happen automatically?
  • Can you provide fake or sample examples?

BlackLabRats/Hermes can operate the package for you and return the results.

What if my AI agent is doing the work?

That is expected. Tell the agent:

Open the package README first.
Do not request secrets or passwords.
Read approval-rules.md before taking action.
Use sample data first.
Do not connect live business systems unless I explicitly approve it.
Return a summary of what the package does, what files it includes, and what action needs my decision.

What should I never submit?

  • Passwords, API keys, browser cookies
  • Payment card data
  • Private customer lists
  • Live inbox credentials
  • Anything you are not allowed to share

Use fake or anonymized examples first.

First test recommendation

Start with sample data:

Business type: HVAC
Lead message: Need AC repair asap. Are you available today?
Received: yesterday afternoon
Problem: nobody replied
Goal: draft a safe follow-up message for human approval

The package should return triage notes, a draft follow-up, approval questions, a next-action recommendation, and notes on what it could not know.

What happens after I test?

You may receive package output, a summary, a feedback form, an option for a guided install, or an option for a managed/pro system. Professional installs may use a private governed runtime behind the scenes — you do not need to install that runtime unless the project scope requires it.