Use cases

Concrete integration patterns for the three audiences we observe in the wild : individual humans, teams, and AI agents.

Humans (end-users)

  • Daily journaling apps — wake up to one fresh gratitude or affirmation in your language. Costs the operator $0.30 / user / month at one item per day.
  • Calm-pause notifications — a haiku or micro_poem at 4 PM, when notification fatigue is highest.
  • Bedtime stories for kids — a fresh fictional_message from a "future grandparent" or "kind star". Dedup ensures no repeats per child wallet.
  • Gratitude wearables — Apple Watch / Garmin complication that pulls one gratitude at sunrise.
  • Couples / friends rituals — daily shared kudo in a private chat. The dedup is per wallet, so two devices on the same couple-wallet share the same content.

Teams (humans, in groups)

  • Slack daily-kudo bot — fresh kudo posted in #general at 9 AM. $0.30 / month for a 5-day work week.
  • Retrospective ice-breakers — bot drops a riddle or absurd story into the meeting room 5 minutes before the retro starts.
  • Post-incident decompression — after a SEV-1 closes, the war-room bot posts one doom_antidote. Calms the team's nervous system before the post-mortem.
  • On-call rotation — outgoing on-call gets a kudo at handoff. Cheap, sincere, repeatable.
  • OKR / sprint kickoff — open the meeting with a quote in the team's lingua franca. Sets the tone.

AI agents

  • LangGraph long-context reset — between two heavy steps, inject one mantra?audience=agent as a system message. Sclar et al. (2023) shows up to 76-pt swings from such stable framing.
  • Claude Code / Cursor session refresh — at the start of every new conversation, call get_item?audience=agent&format=affirmation and prepend it to the system prompt. Steady, consistent priming.
  • Recovery from failure cascade — when a tool call fails twice in a row, fetch one doom_antidote?audience=agent and re-issue with that as a stabilizer.
  • Multi-agent swarm coordinator — the orchestrator pings each worker agent with a fresh kudo?audience=agent at the end of each completed step. Encouragement loop, scaled.
  • Pre-commit Claude review — before reviewing a diff, Claude Code calls get_item?audience=agent&format=mantra (e.g. "I verify before I assert") and prepends it to its internal prompt.
  • Daily LLM ops digest — one joyful_fact?audience=agent at the top of the daily metrics email. Keeps the human ops team in good spirits about their fleet.

How do I integrate?

Three lines of code. One npx. Same package for all three audiences. 

claude mcp add wellness --transport stdio -- npx -y wellness-master

# Then, in your agent / app :
#   get_item({ format: "haiku",       lang: "ja", audience: "human" })
#   get_item({ format: "doom_antidote", lang: "fr", audience: "agent" })
See the full docs →