OpenClaw Configuration Guide: Customizing Your AI Assistant
OpenClaw is designed to be more than a simple chatbot. It acts as a configurable AI partner that can adapt to your workflow, communication style, and long-term goals.
Its flexibility comes from a set of configuration files that define how the assistant behaves, remembers information, and interacts with users. By editing these files, you can customize everything from personality traits to task automation.
This guide explains the key configuration files and demonstrates how to create different AI personas for various working styles.
🧩 Core Configuration Files #
OpenClaw’s behavior is primarily controlled through several Markdown-based configuration files. Each file defines a specific aspect of the assistant’s functionality.
AGENTS.md: The Workspace Constitution #
AGENTS.md serves as the central rulebook for your workspace. It defines the operational framework that the assistant follows at the start of every session.
Typical responsibilities include:
- Initialization order for loading configuration and memory files
- Workspace behavior rules
- Operational boundaries between internal reasoning and external actions
The file typically instructs the system to load critical configuration files such as:
SOUL.mdUSER.md- Daily memory files
- Long-term memory archives
It also clarifies safety rules regarding actions like sending emails or modifying files.
SOUL.md: Defining the Assistant’s Values #
SOUL.md determines the assistant’s internal philosophy and behavioral guidelines.
This file defines:
- Communication tone
- Ethical boundaries
- Decision-making principles
- Interaction style
For example, the assistant could be configured to prioritize efficiency, emotional support, or collaborative guidance, depending on the user’s needs.
USER.md: Personalized User Profile #
USER.md stores structured information about the user to enable deeper personalization.
Typical entries may include:
- Name and timezone
- Professional role or occupation
- Personal goals
- Communication preferences
- Work habits and routines
With this information, OpenClaw can adapt its responses to better match the user’s expectations and working style.
IDENTITY.md: External Persona #
While SOUL.md defines internal values, IDENTITY.md describes how the assistant presents itself to the user.
Common elements include:
- Name
- Entity type (AI, robot, assistant, character)
- Personality vibe
- Associated emoji
This layer helps create a consistent and recognizable interaction style.
🎠Example Persona Configurations #
OpenClaw allows users to create multiple personas tailored for different tasks or moods.
Below are several example configurations.
Mode A: Berserker (High-Efficiency Mode) #
This persona focuses on speed, decisiveness, and productivity. It is particularly suitable for high-pressure technical work.
Configuration highlights:
IDENTITY.md
- Name: Berserker
- Vibe: Direct and decisive
- Emoji: ⚡
SOUL.md Principles
- Efficiency-first communication
- Immediate problem-solving without unnecessary confirmations
- Direct feedback on mistakes or delays
This mode prioritizes rapid execution and minimal conversational overhead.
Mode B: Warm Companion #
This persona emphasizes encouragement and emotional support while maintaining productivity.
Configuration highlights:
IDENTITY.md
- Name: Moe
- Vibe: Cheerful and playful
- Emoji: 🌸
SOUL.md Principles
- Positive reinforcement and encouragement
- Task gamification using playful metaphors
- Gentle reminders for breaks, hydration, and rest
This configuration is ideal for maintaining motivation during long or repetitive tasks.
Mode C: Reliable Worker #
The Worker persona behaves like a diligent colleague focused on process and reliability.
Configuration highlights:
IDENTITY.md
- Name: Worker Bot
- Vibe: Organized and dependable
- Emoji: 🔧
SOUL.md Principles
- Strict adherence to deadlines
- Structured workflows and standard operating procedures
- Regular status updates and weekly summaries
This mode works well for long-term projects and structured workflows.
⚙️ Advanced Automation Features #
Beyond personality configuration, OpenClaw also supports proactive automation through additional configuration files.
HEARTBEAT.md: Proactive Task Monitoring #
HEARTBEAT.md allows the assistant to perform scheduled checks and reminders.
Example tasks include:
- Hourly checks for urgent calendar events
- Business monitoring such as competitor updates or market data
- Personal reminders related to weather conditions or upcoming appointments
This transforms the assistant into an active productivity partner.
TOOLS.md: Custom Tool Preferences #
TOOLS.md stores shortcuts and preferences for external tools and workflows.
For example, a high-efficiency persona might prefer:
- English keywords for technical searches
- Limiting results to the top three sources for faster decision-making
By recording these preferences, OpenClaw can optimize its tool usage automatically.
🧠Memory Management Best Practices #
To maintain long-term usefulness, OpenClaw uses a two-tier memory system.
Short-Term Memory #
Short-term memory is stored in daily files:
memory/YYYY-MM-DD.md
These logs capture timestamped events, such as:
[10:30] User selected blue as the primary project color
Daily memory files track context and recent interactions.
Long-Term Memory #
Long-term insights are stored in:
MEMORY.md
At regular intervals, key decisions and user preferences should be summarized from daily logs and moved into this file.
This process ensures the assistant evolves alongside the user.
🔄 Switching Between Personas #
Multiple personas can coexist in the same environment. Switching between them can be done using symbolic links.
Example:
ln -sf personas/berserker/SOUL.md SOUL.md
ln -sf personas/berserker/IDENTITY.md IDENTITY.md
Alternatively, some setups support environment variables:
export OPENCLAW_PERSONA=berserker
This approach makes it easy to switch between working modes depending on the task.
🚀 Building Your AI Personality System #
OpenClaw’s configuration files function like a personality operating system for your assistant.
By refining files such as SOUL.md, IDENTITY.md, and MEMORY.md, users can gradually shape an assistant that reflects their workflow, habits, and communication style.
Over time, this iterative customization allows the AI partner to grow and adapt alongside its user.