Introduction

Get Started

Install

playbooks.cool installs locally on Mac. Download the app, move it into Applications, and open the workspace where the work already lives. The first run keeps setup intentionally boring: no package manager, no prompt library, no terminal commands for non-technical teammates.

1. Download the desktop app and open it from Applications or the Start menu.
2. Create or join a workspace for your team.
3. Choose a pack that matches the job: compliance, pharmacy, legal, or research.
4. Run a starter playbook against a safe folder, export, or document set.

Get Started

Your first playbook

Start with a workflow that already has a review habit. Good first candidates have clear inputs, a known definition of done, and a person who can approve the result.

Get Started

Invite your team

Invite teammates once the first playbook is clear enough to run without you in the room. The goal is not to teach them prompt engineering. The goal is to give them a named workflow with inputs, expected output, and a review path.

• Give each playbook a job name the team already uses.
• Write input notes in business language, not tool language.
• Make the approver explicit before the playbook can write back.
• Keep early packs small so people can find the right workflow quickly.

Concepts

Playbooks

A playbook is a reusable AI workflow. It packages the goal, input expectations, working rules, review requirements, and output format into something a teammate can run without copying prompts into a chat box.

Playbooks work best when the task has a repeated shape: quality control, purchasing comparison, legal drafting, research synthesis, reporting, or deadline checks. A playbook can still ask clarifying questions, but it should not require the teammate to know model settings, shell commands, or repository conventions.

Concepts

Packs

Packs group related playbooks around a team workflow. A compliance pack might include policy checks, evidence collection, remediation drafts, and sign-off summaries. A pharmacy pack might include purchasing analysis, substitution checks, and dashboard updates.

Concepts

Review packets

A review packet is the handoff between AI work and human approval. It should make the result inspectable: what changed, why it changed, what sources were used, which assumptions remain, and what needs approval.

Concepts

Safe copies

The app works on safe copies by default. That means the AI can explore, draft, compare, and test without silently mutating the original workspace. The original file, folder, or export remains the reference point until a person approves the packet.

This pattern is especially useful when the inputs are operational: contracts, procurement exports, case research, policies, dashboards, or evidence folders. The team can inspect the proposed output before deciding what belongs in the real system.

Concepts

Approvals

Approvals are the boundary between suggestion and action. A playbook can prepare a draft, identify missing evidence, and propose updates, but sensitive changes should wait for a person with the right responsibility.

• Use approval steps for source-of-truth writes, external sends, and client data.
• Route domain decisions to the person who already owns that decision today.
• Keep approval language concrete: accept, reject, edit, or rerun with notes.

Reference

Workspace setup

A workspace should mirror how the team talks about the work. Use packs for business areas, playbooks for repeated jobs, and review packets for decision handoff.

Reference

Models

Model choice should be invisible to most teammates. Workspace owners can choose the model profile that fits the pack: faster runs for simple extraction, stronger reasoning for legal or compliance review, and stricter review expectations for high risk work.

When in doubt, make the playbook output easier to review instead of asking users to tune the model. Better packets beat clever settings.

Reference

Data and privacy

playbooks.cool is designed around local work and reviewable handoffs. The app runs on Mac, keeps the user close to the files they selected, and treats approvals as explicit actions rather than background automation.

For sensitive workflows, keep pack instructions narrow, use safe copies, avoid adding unneeded folders, and require a review packet before exporting or writing results back to another system.

Reference

Team workflow

Roll playbooks out like operating procedures. Start with a real workflow, keep one owner accountable for the first version, and improve the packet every time a reviewer has to ask the same question twice.

1. Pick the repeated job and define the review owner.
2. Create the smallest playbook that gets to a useful packet.
3. Run it on two or three known examples and compare reviewer effort.
4. Save it into a pack only when the team vocabulary feels natural.

Reference

Troubleshooting