Workflows & Recipes

What you need

• OpenClaw running
• Telegram or WhatsApp connected
• Gmail skill (optional)
• Google Calendar skill (optional)

Steps

1. 1Open your OpenClaw chat (Telegram or WhatsApp).
2. 2Send: "Set up a morning briefing cron that runs at 7am every day."
3. 3The agent will ask which data sources you want — pick from weather, calendar, email, to-dos.
4. 4Approve the cron configuration when prompted.
5. 5Done — you'll get a formatted digest every morning.

The prompt / config

Schedule a daily cron at 7:00 AM Pacific that:
1. Checks today's weather for San Mateo, CA
2. Lists my Google Calendar events for today
3. Summarizes my top 5 unread Gmail messages
4. Sends the combined briefing to this chat

Format it cleanly with headers and emoji.

Expected result

Every morning at 7 AM, you get a single message with weather, schedule, and email highlights — no apps to open.

What you need

• OpenClaw running
• Telegram or WhatsApp connected

Steps

1. 1Tell the agent your current goals (e.g., fitness, project milestones, habits).
2. 2Ask it to set up a daily check-in cron at your preferred time.
3. 3The agent stores your goals in a memory file and asks about each one daily.
4. 4Reply with updates — it tracks streaks and progress over time.

The prompt / config

Set up a daily accountability check-in at 9 PM. My current goals:
1. Work out 5x/week
2. Ship one feature per day on my side project
3. Read for 30 minutes before bed

Every evening, ask me about each goal. Track my streak in a memory file. If I miss 2 days in a row, call me out.

Expected result

A persistent accountability partner that doesn't forget, doesn't let you off the hook, and tracks your streaks.

What you need

• OpenClaw running
• Gmail skill
• ElevenLabs API key (for TTS)
• Telegram connected

Steps

1. 1Connect your Gmail account to OpenClaw.
2. 2Set up an ElevenLabs API key in your environment.
3. 3Ask the agent to create a morning cron that reads your top emails as a voice message.
4. 4The agent fetches unread emails, summarizes them, converts to speech, and sends as a voice note.

The prompt / config

Create a morning cron at 7:30 AM that:
1. Fetches my 10 most recent unread Gmail messages
2. Summarizes each in one sentence
3. Converts the full summary to a voice message using ElevenLabs
4. Sends the voice note to this chat

Use a natural, conversational tone — like a radio host reading the news.

Expected result

A voice note in your chat every morning summarizing your inbox — listen while brushing your teeth or commuting.

What you need

• OpenClaw running
• Telegram or WhatsApp connected

Steps

1. 1Send the agent a product URL and your target price.
2. 2It sets up a periodic check (every 4-6 hours) that scrapes the page.
3. 3When the price drops below your target, you get a notification.

The prompt / config

Monitor this product for price drops:
URL: https://www.amazon.com/dp/B0EXAMPLE
Target price: $49.99

Check every 6 hours. When the price drops below my target, message me immediately with the current price and a direct link.

Expected result

Hands-free price tracking — get pinged the moment a deal hits your target.

What you need

• OpenClaw running
• Gmail OAuth connected
• Telegram connected

Steps

1. 1Connect Gmail to OpenClaw via OAuth.
2. 2Tell the agent to audit your inbox — it categorizes senders by type (marketing, transactional, personal, important).
3. 3Review and approve the unsubscribe list it generates.
4. 4The agent bulk-unsubscribes, archives noise, and surfaces what matters.
5. 5Set up a daily triage cron to keep it clean going forward.

The prompt / config

Audit my Gmail inbox:
1. Categorize all senders: marketing, transactional, personal, important
2. Generate a list of newsletters/marketing to unsubscribe from — show me first
3. After I approve, unsubscribe and archive those threads
4. Surface my top 10 most important unread messages
5. Set up a daily cron at 8 AM to auto-triage new mail

Rules: Never auto-delete. Always ask before unsubscribing. Star anything from a real person.

Expected result

10,000 emails cleared on day one. Ongoing inbox stays under 20 messages with daily auto-triage.

What you need

• OpenClaw running
• Google Calendar skill
• Web search access
• Telegram connected

Steps

1. 1Connect your Google Calendar to OpenClaw.
2. 2Ask the agent to set up a pre-meeting research cron.
3. 3It checks your calendar 30 minutes before each meeting.
4. 4For each attendee, it searches LinkedIn, recent posts, company news, and shared connections.
5. 5Delivers a formatted briefing doc to your chat before the meeting starts.

The prompt / config

Set up a pre-meeting briefing system:
1. Check my Google Calendar every 15 minutes for upcoming meetings
2. 30 minutes before each meeting, research every attendee:
   - LinkedIn profile summary
   - Recent public posts or articles
   - Company/role info
   - Any shared connections or past interactions
3. Send me a formatted briefing in this chat

Skip recurring 1:1s with people I meet weekly. Focus on new contacts and important meetings.

Expected result

Walk into every meeting prepared. No more scrambling to Google someone 2 minutes before a call.

What you need

• OpenClaw running
• WhatsApp or Signal connected
• Telegram for digest delivery

Steps

1. 1Connect the group chat you want summarized.
2. 2Tell the agent which group(s) to monitor and when to send the digest.
3. 3It reads messages throughout the day, identifies topics, decisions, and action items.
4. 4At your chosen time, sends a structured summary to your private chat.

The prompt / config

Monitor my WhatsApp group "Startup Founders" and send me a daily digest at 6 PM:
1. Key topics discussed today
2. Decisions made (who decided what)
3. Action items (who owes what)
4. Interesting links shared
5. Any questions directed at me

Keep it under 500 words. Skip small talk and reactions.

Expected result

Stay informed on busy group chats without reading 200+ daily messages. Never miss a decision or action item.

What you need

• OpenClaw running
• GitHub CLI configured
• Coding agent (Claude Code/Codex)
• Telegram connected

Steps

1. 1Make sure your GitHub repo is accessible and `gh` CLI is authenticated.
2. 2Send the agent a feature request or bug description via chat.
3. 3It creates a GitHub issue, spawns a coding sub-agent in a worktree.
4. 4The sub-agent implements the change, runs tests, and opens a PR.
5. 5You review and merge from your phone.

The prompt / config

I need a new API endpoint in my Express app:

POST /api/v1/feedback
- Accepts: { rating: 1-5, comment: string, email?: string }
- Validates input, stores in SQLite
- Returns 201 with the created record
- Add tests

Create a GitHub issue, implement it, and open a PR to the main branch.
Repo: my-org/my-api

Expected result

Ship features from anywhere — describe what you want, get a tested PR ready for review.

What you need

• OpenClaw running
• Home Assistant or Philips Hue
• Google Calendar (optional)
• Telegram connected

Steps

1. 1Connect OpenClaw to your smart home system (Home Assistant API or direct Hue bridge).
2. 2Let the agent discover your devices — it scans the network automatically.
3. 3Define your morning routine: which lights, what brightness, any music.
4. 4Add calendar and weather to the routine for a full morning experience.
5. 5Set up a cron or trigger it manually with a command like "Good morning."

The prompt / config

Set up a "Good morning" routine:
1. Turn on living room lights to 60% warm white
2. Turn on kitchen lights to 80%
3. Start playing "Morning Jazz" on the kitchen HomePod
4. Check today's weather and calendar
5. Send me a brief summary: weather, first meeting time, and any reminders

Trigger: cron at 6:45 AM weekdays, or when I say "Good morning" in chat.

Expected result

Your home wakes up with you — lights, music, and a briefing, all hands-free.

What you need

• OpenClaw running
• Multiple model API keys (Opus, Sonnet, Haiku)
• GitHub access
• Telegram connected

Steps

1. 1Plan your agent roles: Strategy (high-level planning), Dev (coding), Marketing (content + social), Business (ops + finance).
2. 2Create separate OpenClaw agent configs with role-specific system prompts.
3. 3Assign models by role: Opus for strategy, Sonnet for dev and marketing, Haiku for routine business ops.
4. 4Set up shared memory files so agents can reference each other's output.
5. 5Configure each agent with appropriate tool access (GitHub for dev, social APIs for marketing, etc.).
6. 6Test each agent independently, then run them in parallel.

The prompt / config

I'm setting up a multi-agent system for my solo SaaS business. Create configs for:

1. **Strategy Agent** (Opus) — weekly planning, competitive analysis, roadmap
2. **Dev Agent** (Sonnet) — feature implementation, bug fixes, deployments
3. **Marketing Agent** (Sonnet) — content drafts, social posts, email campaigns
4. **Business Agent** (Haiku) — invoice generation, metrics tracking, customer support triage

Shared memory: ~/agents/shared/ (company context, product info, customer data)
Individual memory: ~/agents/{role}/ (role-specific context)

Each agent should be able to read shared memory but only write to their own directory.

Expected result

A virtual team that operates in parallel across your business — each agent specialized, all sharing context.

What you need

• OpenClaw running
• BugSnag API key
• GitHub CLI
• CI/CD pipeline
• Coding agent

Steps

1. 1Connect BugSnag (or Sentry) API to your OpenClaw environment.
2. 2Set up a nightly cron that pulls new/unresolved errors.
3. 3For each error, the agent: reproduces the issue, writes a regression test, implements a fix.
4. 4It runs the full test suite to ensure no regressions.
5. 5Opens a PR with the fix and test — or auto-deploys if configured.

The prompt / config

Set up a nightly bug-fix cron at 2 AM:
1. Pull new errors from BugSnag API (last 24 hours, severity: critical + major)
2. For each error:
   a. Analyze the stack trace and identify the root cause
   b. Write a regression test that reproduces the issue
   c. Implement the minimal fix
   d. Run the full test suite
   e. Open a PR with title "fix: [BugSnag #{id}] {error_message}"
3. Send me a morning summary of what was fixed

Skip errors tagged as "won't fix" or "expected". Max 5 fixes per night.
Repo: my-org/my-app

Expected result

Wake up to fewer bugs and more green builds. Routine errors fixed overnight without human intervention.

What you need

• OpenClaw running
• Web search access
• Social media API keys (optional)
• Telegram connected

Steps

1. 1Define your content strategy: topics, tone, target audience, posting schedule.
2. 2The agent researches trending topics in your niche using web search.
3. 3It drafts content matching your voice (learned from past posts).
4. 4You review and approve (or request edits) via chat.
5. 5Approved content gets scheduled across your platforms.

The prompt / config

Set up a weekly content pipeline for my technical blog:

1. **Monday**: Research trending topics in AI agents / developer tools. Send me 5 topic ideas with briefs.
2. **Tuesday**: After I pick a topic, research it deeply — find data, examples, quotes.
3. **Wednesday**: Draft a 1500-word blog post in my voice (direct, technical, slightly opinionated).
4. **Thursday**: Send me the draft for review. Incorporate my feedback.
5. **Friday**: Publish to my blog and create social posts (X thread + LinkedIn post).

Store my content calendar in ~/content/calendar.md. Track what performs well.

Expected result

Consistent content output without the blank-page paralysis. One review session per week produces a full blog post + social posts.

What you need

• OpenClaw running
• Web browsing capability
• Vision/image analysis
• Telegram connected

Steps

1. 1Define your search criteria in natural language — including nuanced rules the agent needs to reason about.
2. 2Provide example listings that match and don't match your criteria.
3. 3The agent sets up periodic monitoring of listing sites.
4. 4For each new listing, it analyzes photos, reads descriptions, and applies your reasoning criteria.
5. 5Matches get sent to your chat with a summary of why they qualified.

The prompt / config

Monitor vacation rentals in Maui for my trip in March:

Criteria:
- Budget: $150-250/night
- Must have: ocean view, kitchen, WiFi, AC
- Bedrooms: 2+, but a pullout bed is OK if it's NOT in the same room as another bed
- Pool: preferred but not required. Shared pool is fine if reviews mention it's not crowded.
- NO ground floor units (safety concern)
- Check photos: reject if furniture looks worn or bathroom looks dated

Check Airbnb and VRBO every 4 hours. For each match:
- Show top 3 photos
- Price per night
- Why it matches my criteria
- Any yellow flags from reviews
- Direct booking link

Expected result

Find exactly what you're looking for without endless scrolling. The agent applies your taste and reasoning to every listing.

What you need

• OpenClaw running
• Gmail
• Google Calendar
• WHOOP/health wearable (optional)
• Task manager (Linear, Notion, etc.)
• Telegram connected

Steps

1. 1Connect all your data sources: email, calendar, health wearable, task manager, news feeds.
2. 2Define the sections you want in your dashboard and their priority order.
3. 3Set up a morning cron that aggregates everything into one structured briefing.
4. 4Optionally add an evening cron for a daily wrap-up and next-day preview.

The prompt / config

Create a comprehensive daily dashboard cron at 6:30 AM:

## Health (WHOOP)
- Recovery score, HRV, sleep quality, strain from yesterday
- Flag if recovery is below 33% (suggest rest day)

## Calendar
- Today's meetings with prep notes
- Tomorrow's first meeting (so I can prep tonight)

## Tasks
- Top 5 priority tasks from Linear
- Any overdue items (call them out)

## Email
- Unread count + top 3 important emails summarized

## News & Markets
- S&P 500 pre-market, any >2% moves in my watchlist
- Top 3 AI/tech news stories

## Weather
- Today + tomorrow, highlight if rain or extreme temps

Format as a single clean message with emoji headers. Keep total under 800 words.
Evening cron at 9 PM: short wrap-up of what I completed + tomorrow preview.

Expected result

One message replaces checking 6+ apps every morning. Full situational awareness in 60 seconds of reading.