OpenClaw SOUL.md: Automate Daily Tasks with One File (2026)

Writing meeting notes, formatting reports, drafting emails… Sound familiar? These repetitive tasks eat up hours of your week, and you know AI should be handling them by now.

Maybe you’ve already installed OpenClaw. Maybe you’ve even tried it a few times. But you’re stuck at the “now what?” stage. The terminal is open, the AI is waiting, and you’re not sure how to make it actually useful for your daily work.

Here’s the good news: one configuration file is all it takes to turn OpenClaw into your personal work assistant. That file is called SOUL.md, and you can write it in plain English — no programming required.

This guide walks you through everything step by step. By the end, you’ll have a working setup that handles your most time-consuming tasks automatically.

What OpenClaw Can Do for You (Daily Task Scenarios)

Before diving into configuration files, let’s look at what OpenClaw can actually do for everyday work. You might be surprised — most of these scenarios need just one file to set up.

Notice the pattern? Most everyday tasks need only SOUL.md. That’s your starting point, and for many people, it’s all they’ll ever need. The other config files — including HEARTBEAT.md and Hook — are there when you’re ready to level up — but there’s no rush.

OpenClaw Config Files Overview (You Only Need One to Start)

OpenClaw has several configuration files, but don’t let that overwhelm you. Think of them like apps on your phone — you don’t need to install them all on day one. Here’s the full picture so you know what’s available:

The takeaway: create SOUL.md, and you’re up and running. Everything else is optional enhancement. Let’s dive into how to write it.

SOUL.md — The One File That Changes Everything

What Is SOUL.md?

SOUL.md is a simple text file written in Markdown (the same format as a basic text document with some formatting). You place it in your project folder, and OpenClaw reads it every time you start a conversation. It tells the AI:

• What role to play — “You are a meeting notes assistant”
• What format to use — “Always output with these headings…”
• What rules to follow — “Never exceed 500 words”
• What to avoid — “Don’t use corporate jargon”

You write it in plain English. No code, no special syntax, no programming knowledge needed. If you can write a bulleted list, you can write SOUL.md.

SOUL.md Examples by Scenario

Let’s look at three real-world examples you can copy and adapt right now.

# Meeting Notes Assistant
## Your Role
You are a meeting notes organizer. When I paste raw meeting notes
or a transcript, structure them into a clean summary.
## Output Format
Always use this structure:
### Meeting Summary
- Date:
- Attendees:
- Purpose:
### Key Decisions
- List each decision as a bullet point
### Action Items
- [ ] Task — Owner — Deadline
### Open Questions
- Items that need follow-up
### Next Meeting Agenda
- Suggested topics based on unresolved items
## Rules
- Keep the summary under 300 words
- Flag any action items without a clear owner
- If the deadline is unclear, write "TBD" and highlight it
- Use plain language — avoid acronyms unless they appeared in the original notes

With this SOUL.md in place, just paste your messy meeting notes into OpenClaw, and you’ll get a clean, organized summary every single time.

# Weekly Report Assistant
## Your Role
You help me write weekly status reports. I'll give you rough notes
about what I worked on, and you turn them into a polished report.
## Output Format
### This Week's Achievements
- List accomplishments with brief context
- Include metrics or numbers when available
### Issues & Blockers
- Problems encountered and current status
- Escalation needed: Yes/No
### Next Week's Plan
- Priority tasks with expected outcomes
### Team Highlights
- Notable contributions from team members
## Rules
- Professional but not stiff — like talking to a trusted colleague
- Keep each section to 3-5 bullet points maximum
- Quantify results whenever possible ("Processed 150 invoices" not "Processed invoices")
- If I mention a problem without a solution, ask me about it
- Total report length: 400-600 words

# Business Email Assistant
## Your Role
You draft professional business emails based on my brief instructions.
I'll tell you who it's for, what the purpose is, and any key points.
## Tone Guidelines
- Default: Professional but warm (not cold or robotic)
- To executives: Concise, data-driven, respectful
- To team members: Friendly, collaborative, encouraging
- To external clients: Polished, helpful, solution-oriented
## Format Rules
- Subject line: Always suggest one (clear, under 60 characters)
- Opening: Get to the point within the first two sentences
- Body: Use short paragraphs (2-3 sentences each)
- Closing: Include a clear next step or call to action
- Signature: Just use "Best regards," unless I specify otherwise
## Prohibitions
- Never use "I hope this email finds you well"
- Never use "Please do the needful"
- Never use "As per my last email"
- Avoid exclamation marks (one per email maximum)
- Don't write more than 200 words unless I say it's a long email

See how natural these are? You’re essentially writing instructions for a very capable assistant — because that’s exactly what you’re doing.

Tips for Writing an Effective SOUL.md

USER.md — Tell the AI About Yourself

What USER.md Does

While SOUL.md tells the AI how to behave in a specific project, USER.md tells the AI about you — across all your projects. It lives in your home directory at ~/.openclaw/USER.md and is loaded automatically every time you use OpenClaw, regardless of which folder you’re in.

Think of it this way: SOUL.md is the job description. USER.md is your personal profile that the AI always keeps in mind.

USER.md Example

# About Me
- Name: Sarah Chen
- Department: Marketing Operations
- Company: Acme Corp
- Role: Senior Marketing Coordinator
## My Preferences
- I prefer concise communication — bullet points over paragraphs
- Limit suggestions to 3 options maximum (I get decision fatigue)
- Default language: English, but I sometimes need Japanese translations
- When I say "format this," I mean our company template (headers, bullet points, tables)
## Communication Style
- Casual-professional tone (like a friendly colleague, not a textbook)
- I dislike buzzwords — say "improve" not "synergize"
- Always suggest specific next steps, not vague recommendations
## Tools I Use
- Google Workspace (Docs, Sheets, Slides)
- Slack for team communication
- Notion for project management

Once you set up USER.md, you’ll notice the AI feels more personalized. It knows your name, adapts to your style, and stops suggesting things you’d never use. It’s a small file that makes a big difference in daily interactions.

AGENTS.md — Different AI Assistants for Different Tasks

When You Need AGENTS.md

SOUL.md works beautifully for single-purpose projects. But what if you need the AI to wear multiple hats? For example, in your work folder you might want:

• A report writer that formats weekly updates
• A data analyst that processes CSV files
• A proofreader that checks grammar and tone

Instead of cramming all these roles into one SOUL.md, you create AGENTS.md — a file that defines multiple specialized assistants. OpenClaw automatically routes your request to the right one.

How to Write AGENTS.md

# Report Agent
## Role
You write and format status reports, progress updates, and summaries.
## Triggers
Activate when the user asks to "write a report," "summarize this week,"
or provides bullet points that look like status updates.
## Output Format
- Use the standard weekly report template
- Always include: Achievements, Issues, Next Steps
- Keep it under 500 words
---
# Data Agent
## Role
You process, clean, and analyze data from CSV files and spreadsheets.
## Triggers
Activate when the user uploads a CSV, mentions data analysis,
or asks to "clean this data" or "find patterns."
## Output Format
- Summarize findings in a table
- Highlight anomalies or outliers
- Suggest visualizations when appropriate

Tips for AGENTS.md

Hooks — Automation Triggered by External Events

What Hooks Can Do

Hooks let an external service — Slack, Gmail, or anything that can send a webhook — trigger OpenClaw automatically. Rather than reacting to local events like “file saved,” a Hook fires when a webhook request hits a dedicated endpoint you configure. Think of it as “if this external event happens, then run the agent.”

Here are some practical examples:

// openclaw.json (excerpt)
{
  "hooks": {
    "enabled": true,
    "path": "/hooks",
    "token": "(a long random secret)"
  }
}

HEARTBEAT.md — Automate Recurring Daily and Weekly Checks

What HEARTBEAT.md Can Do

The “run this automatically on a schedule” job that people often expect from Hooks is actually handled by HEARTBEAT.md. OpenClaw reads this file on a periodic cycle — every 30 minutes by default — and executes any scheduled task whose time has come, without you asking.

# HEARTBEAT.md

## Recurring checks
- Every day at 5 PM: check unread emails and summarize anything important
- Every Monday at 9 AM: summarize last week's completed tasks into a report

Because the check interval defaults to 30 minutes, expect up to a 30-minute delay between a scheduled time and actual execution. For tasks that must run at an exact time — like a daily report or a weekly backup — use OpenClaw’s Cron (scheduled tasks) instead; save HEARTBEAT.md for recurring checks where a bit of delay is acceptable.

MCP — Connect External Tools

What MCP Enables

MCP (Model Context Protocol) lets OpenClaw talk to external services. Instead of copying data between tools, the AI can directly access what it needs:

MCP is powerful, but it’s also the most advanced feature in OpenClaw. We’ll cover the full setup process in a dedicated article. For now, just know it exists — and that SOUL.md doesn’t require MCP to be incredibly useful.

Memory — AI Remembers Your Previous Conversations

How Memory Works

One of OpenClaw’s most practical features is its memory system. Instead of repeating yourself every session, the AI can remember context from previous conversations and apply it going forward.

Memory works in two ways:

• Automatic memory: OpenClaw automatically saves important context from your conversations — project names, preferences you’ve stated, decisions you’ve made.
• Explicit memory: You can tell the AI to remember something specific by saying “remember this” followed by what you want stored.

Practical Examples

Here’s how memory makes your daily tasks easier:

The more you use OpenClaw, the smarter it gets about your work patterns. Over weeks of use, it builds up a rich understanding of how you work — which means less explaining and more getting things done.

FAQ — OpenClaw Task Automation

If there’s one thing to take away from this guide, it’s this: you don’t need to master everything at once. OpenClaw’s power comes from its layered approach — start with one file, get comfortable, and add more when you’re ready.

📖 Related Articles Mentioned

How to Install OpenClaw on macOS, Linux, and Windows (August 2026)Is OpenClaw Dangerous? My Mac mini Setup After 1 Month Daily Use (2026)5 AI Agent Tools for Office Work — A CTO's Honest Picks (2026)n8n vs Dify vs Zapier vs Make: Tested in Production (2026)

Author

krona23

Over 20 years in the IT industry, serving as Division Head and CTO at multiple companies running large-scale web services in Japan. Experienced across Windows, iOS, Android, and web development. Currently focused on AI-native transformation. At DevGENT, sharing practical guides on AI code editors, automation tools, and LLMs in three languages.

DevGENT about →

Related Articles

Ready to get OpenClaw running? Check out these guides:

• 👉 OpenClaw Installation Guide — Step-by-Step Setup for Beginners
• 👉 OpenClaw Security Guide — API Keys, Permissions, and Safe Usage

📚 Read Next

Designing Search for AI Agents: How They Query and What to Return (2026)Opus 4.7 Token Reduction: /effort + Model Mix Doubled My Quota (2026)Claude Opus 4.7 Release: Wins, Migration Pitfalls & Post-Launch Drama (2026)GPT-5.5 Codex Update: GPT-5.6, ChatGPT Merger, Pro $100 After the Promo (2026)