Connect and run your first loop
Connect the MCP
Three ways in, pick one.
Claude Code, inside this repo. The repo ships a .mcp.json pointing at https://xapi.getstanley.ai/mcp. Open the cloned folder in Claude Code, approve the project MCP server when asked, then run /mcp to sign in.
Claude Code, any folder.
claude mcp add --transport http stanley https://xapi.getstanley.ai/mcp
Then /mcp to sign in.
claude.ai. Settings → Connectors → Add custom connector, paste the URL above.
Audit what's connected
Paste this into Claude:
Ask Stanley to list every platform and integration currently connected for this account, flag anything expired or broken, and show the configured posting slots.
You get back a short rundown: connected accounts, anything with an expired session, missing platforms, and your posting slots with their timezone. Fix gaps at getstanley.ai, not here. Inside this repo, /stanley-start runs the same audit and tells you which recipes your setup can run.
First loop: draft, then schedule
- Draft. "Ask Stanley for one post in my voice for [your strongest connected platform] about [something you did this week]." The draft arrives wrapped in a platform tag like
<tweet>, which is normal (see recipe 09). - Push back once. Tell him what reads wrong. He holds voice context between messages, so one round of notes tends to land.
- Schedule rather than publish. "Schedule this into my next open slot." Scheduling is reversible and asks no permission; publishing live is the step that checks with you first.
- Look at the calendar. "Show me my content calendar for the next 7 days."
Every other recipe in this cookbook is a bigger version of that loop.
When it breaks
- No
message_stanleytool: the server wasn't approved or your sign-in lapsed. Re-run/mcp. - "Nothing connected": link a platform at getstanley.ai first. The cookbook can't do that part for you.
- The draft sounds like anyone: your voice profile is thin. Paste 3 to 5 posts you wrote yourself and ask Stanley to update it.
