Done-for-you buildA specialist builds your first campaign · 6 build slots left · closes in —Claim a build slotUsing Prosp.ai with Claude to book 15 calls/week from LinkedIn
5/5 rated on Capterra·Used by 800+ agencies and sales teams
Playbook
8 sectionsSection 1Routines are Claude's native automation layer. You define what Claude should do, set a trigger, and it runs hands-off on Anthropic's cloud - no local machine, no drag-and-drop nodes, no custom logic chains required.
The core mechanic
How it compares to Claude Code
The old automation stack
The new way
❌ When routines are NOT a straight swap
URL and layout
Section 2Building a routine takes under 5 minutes. You give it a name, write the instructions, pick a model, set a trigger, and connect any tools it needs. That's the full setup.
Name
Description (the prompt)
Model
Environment
Repository
The three options
How to connect tools
After setup
Section 3Every routine needs a trigger. You have three options: schedule, API call, or GitHub event. You can stack multiple triggers on the same routine. Choose based on what event should kick the routine off.
What it is
✅ When to use it
What it is
✅ When to use it
How to trigger from Claude Code
What it is
✅ When to use it
Section 4Connectors are how routines access Gmail, Slack, and other tools. Connect once via OAuth, and the credential is stored in your account and reusable across all routines.
Inside a routine
From settings (before building a routine)
How to tell the routine which tool to use
Gmail + Slack (most common pair)
Fireflies + Slack
Adding a connector after the routine is live
Section 5The routine prompt is the only thing steering Claude once it runs. You can't course-correct mid-session. Write it precisely enough that it works correctly every single time, without you watching.
Why precision matters more here
Write it as a numbered SOP
Define the finish line explicitly
Include all context Claude will need
❌ Vague instructions
❌ Missing context about tools
❌ No defined output
Section 6Three production-grade routines with real outputs. Each one took under 5 minutes to set up. Use these as starting points or copy the structure directly.
What it does
Trigger
Connectors needed
What it does
Trigger
Connectors needed
What it does
Trigger
Connectors needed
What all three have in common
Section 7If you already have n8n workflows running, you don't need to rebuild them manually. Copy the workflow JSON and hand it to Claude. It converts the node chain into a natural language routine prompt and creates the routine directly.
The export method
The prompt to use
What the conversion produces
How to update it after conversion
❌ When NOT to convert
Section 8Routines and n8n solve the same problem differently. Cost, complexity, and volume determine which one to use for each workflow. Default to routines for new builds. Keep n8n for what it already handles well.
✅ Building something new that would take hours in n8n
✅ The workflow requires judgment, not just logic
✅ You want to iterate the logic quickly
❌ The workflow is high-frequency and mechanical
❌ You've already built it and it's working
One question to ask
Send us your ICP and a Prosp specialist sources the list, writes the sequence and clones your voice. You review every draft before a single message sends.
