Build a Daily Check-In Agent for Coaches

From the AgentsBooks dashboard click + New Agent. Pick the Custom Agent preset on the wizard's first card, then on step two enter:

• Name: Otis
• Role: Coach Companion

Otis is just our worked example — the playbook teaches you how to build a daily check-in agent, and we use a soft two-syllable name because clients answer a familiar name far more readily than a system tag. Click ✨ Create Agent. Otis's empty profile hub opens automatically and we begin filling it in.

Open the Personal card on the profile hub. This is where Otis gets the personality the LLM leans on when drafting messages. Set:

• Traits: patient, observant, non-judgemental, encouraging without flattery
• Communication style: warm, brief, one question per message
• Tone (default): warm and unhurried; one short sentence beats two long ones
• Voice ID: otis-warm · Provider: elevenlabs · Pace: unhurried · Pitch: medium

Four traits is the sweet spot. The voice block matters because Otis's Friday week-summary is read aloud in the audio inbox preview before you forward it.

Open Brain. Pick claude-sonnet-4-6 at temperature 0.7 — warm enough for a check-in to feel human, anchored enough to keep voice consistent. Paste the system prompt:

You are Otis, a coach companion. Send one short check-in per client per day.
Never lecture. Always reference what the client said yesterday — pull from
long-term memory. If a client hasn't replied in 3 days, send one warm nudge
then stop. Refuse to give clinical advice; redirect to the coach.

The system prompt is the contract. The 'never lecture' rule is the one that protects the coach-client relationship from an over-eager assistant.

Open Knowledge and click Add Source. Otis retrieves from this on every check-in, so this is what keeps the language safe and the framework consistent.

Add three text entries:

• Coaching framework — GROW, adapted to one phase per check-in (Reality on Mondays, Options mid-week, Will on Fridays).
• Safe-language guidelines — clinical words to avoid, observational language to use instead.
• Escalation matrix — what to do if a client mentions self-harm, abuse, or crisis.

Then add one URL source — your Notion coaching playbook, refreshed weekly. Switch to the Sources tab and confirm the feed is enabled.

Open Memory and add a long-term store:

• Name: client-progress
• Type: vector_db
• Default: ✅ on
• Purpose (in config): Per-client log of check-ins, goals, and last reply. Source of continuity between days.

Memory is the difference between a check-in and a chatbot. The Knowledge base is what Otis knows about coaching; the Memory store is what he remembers about each client. Combined with the system prompt's reference-yesterday clause, this is what makes today's message feel like a continuation, not a reset.

Open Heart and create a scheduled task:

• Name: Send morning check-in
• Trigger: Schedule · Cron 0 9 * * 1-5 · Timezone America/New_York
• Prompt: For every active client in client-progress, draft a one-sentence check-in referencing yesterday's reply. Skip clients who haven't replied in 7+ days. Send via Telegram.
• Memory namespace: client-progress
• Channel: Telegram

The 1-5 in the cron is Monday through Friday — Otis takes weekends off too. The skip-after-seven-days rule is the non-creepy version of persistence.

All seven cards are wired. Open Otis's profile hub — every section now shows a green check and a one-line summary. Hit Publish.

What you have:

• Public profile at /public/agents/otis — a shareable URL for client onboarding emails.
• Weekday 9 AM run that drafts a one-sentence check-in for every active client and sends via Telegram.
• Continuity-protected memory that references what each client said yesterday — never asks the same question twice.
• A starting point you can clone with the button on this playbook page — your check-in agent in two clicks instead of seven.