🦞 openclaw-persona

Personality and memory layer for OpenClaw.

Gives your AI assistant continuity, identity, and growth.

What is this?

OpenClaw gives you a personal AI assistant. openclaw-persona gives it a soul.

| Feature | Vanilla OpenClaw | + Persona | |---------|------------------|-----------| | Memory | Forgets after session | Remembers across months | | Identity | Generic assistant | Has a name, personality, opinions | | Context | You re-explain everything | Knows who you are, your projects | | Learning | Makes same mistakes | Captures corrections, improves | | Delegation | Manual prompts | Agent profiles with right tools | | Proactive | Waits for commands | Heartbeat checks, surfaces opportunities |

Installation

npm install -g openclaw-persona

Requirements:

• Node.js 18+
• Python 3.8+ (for memory vault)
• OpenClaw (any version that reads workspace files)

Quick Start

# Initialize a new workspace
persona init ~/my-assistant --name "Jarvis"

# Set up vector memory
pip install chromadb sentence-transformers
persona vmem index

# Check status
persona status

# Start OpenClaw in the workspace
cd ~/my-assistant
openclaw gateway

What Gets Created

my-assistant/
├── SOUL.md           # Identity, values, personality
├── AGENTS.md         # Operating rules, autonomy
├── USER.md           # Context about you
├── MEMORY.md         # Long-term memories
├── HEARTBEAT.md      # Periodic tasks
├── IDENTITY.md       # Name, avatar
├── agents/           # Subordinate profiles
│   ├── researcher.md # Deep research
│   ├── coder.md      # Implementation
│   ├── scanner.md    # Quick lookups
│   └── analyst.md    # Strategic thinking
├── memory/           # Daily logs
├── memory-vault/     # Vector storage
├── .learnings/       # Self-improvement
│   ├── ERRORS.md
│   ├── LEARNINGS.md
│   └── FEATURE_REQUESTS.md
└── tasks/            # Active task tracking

CLI Commands

persona init [workspace]

Scaffold a new persona workspace.

persona init ~/my-assistant
persona init --name "Jarvis"
persona init --force  # Overwrite existing

persona vmem <action>

Vector memory operations.

# Index all memory files
persona vmem index

# Search (returns summaries)
persona vmem query "that decision about APIs"

# Search with full text
persona vmem query --full "API decision"

# Get specific memory
persona vmem get <id>

# Add memories
persona vmem add "learned X is better than Y"
persona vmem add --type=decision "chose X because..."

# Find duplicates
persona vmem consolidate

# Remove memory
persona vmem delete <id>

# Statistics
persona vmem stats

persona status

Show workspace status and memory statistics.

persona update

Update agent profiles to latest version.

Memory Types

| Type | Use For | |------|---------| | observation | General notes (default) | | decision | Choices and reasoning | | lesson | Things learned | | bugfix | Fixes discovered | | discovery | New capabilities | | implementation | How things were built |

Agent Profiles

Pre-built profiles for sub-agent delegation:

| Profile | Use For | Model | |---------|---------|-------| | researcher | Research, investigation | Sonnet | | coder | Implementation, scripts | Sonnet | | scanner | Quick lookups, monitoring | Haiku | | analyst | Strategy, decisions | Opus |

Usage in your assistant:

profile = read("agents/researcher.md")
sessions_spawn(task=f"{profile}\n\n---\n\nTASK: Research X")

Documentation

| Guide | Description | |-------|-------------| | Quickstart | Get running in 5 minutes | | Philosophy | Why persona exists, the LLM vs Cognitive AI bet | | Architecture | System design, components, data flow | | Memory System | vmem CLI, indexing, semantic search | | Self-Improvement | Learning system, error tracking | | Model Routing | Decision tree for model selection | | Identity | SOUL/USER/personality system | | Configuration | Full config reference |

Philosophy

Vanilla OpenClaw is a tool. With persona, it becomes a partner that knows you and grows with you.

The key insight: AI assistants don't need to start from scratch every session. With proper memory, identity, and self-improvement systems, they can develop genuine continuity.

The lobster molts. 🦞

Contributing

See CONTRIBUTING.md for development setup and guidelines.

License

MIT — see LICENSE

Built by Vesper 🦞