OpenClaw Workspace Template

A ready-to-use workspace template for OpenClaw with a 4-layer context memory system, agent specs, and distillation process.

Quick Start (5 minutes)

# 1. Clone into your OpenClaw workspace
git clone https://github.com/janiceassistant216-prog/openclaw-workspace-template.git ~/.openclaw/workspace

# 2. Edit your identity
nano ~/.openclaw/workspace/SOUL.md      # Who is your AI?
nano ~/.openclaw/workspace/USER.md      # Who are you?
nano ~/.openclaw/workspace/IDENTITY.md  # Name, emoji, vibe

# 3. Start chatting
# Your AI reads these files automatically on session start

What's Included

workspace/
├── AGENTS.md              ← Session startup rules + memory system docs
├── SOUL.md                ← AI personality and behavior (customize this!)
├── USER.md                ← About you (customize this!)
├── IDENTITY.md            ← AI name, emoji, avatar
├── MEMORY.md              ← Long-term curated memory (starts empty)
├── HEARTBEAT.md           ← Periodic check tasks (optional)
├── TOOLS.md               ← Local tool notes (cameras, SSH, etc.)
├── memory/
│   ├── README.md          ← Memory system documentation
│   ├── distillation-process.md  ← How raw events become curated memory
│   ├── agents-spec.md     ← 5 example agent roles with full specs
│   ├── daily/             ← Daily event logs (append-only)
│   │   └── .gitkeep
│   ├── projects/
│   │   └── _example/      ← Example project structure
│   │       ├── MEMORY.md
│   │       ├── RUNBOOK.md
│   │       ├── PROPOSALS.md
│   │       └── agents/coordinator/
│   │           ├── SUMMARY.md
│   │           ├── LOOPS.md
│   │           ├── PROPOSALS.md
│   │           └── LOOPS_ARCHIVE.md
│   └── _TEMPLATES/        ← Blank templates for new projects
│       ├── SUMMARY.md
│       ├── LOOPS.md
│       ├── PROPOSALS.md
│       └── LOOPS_ARCHIVE.md
└── .gitignore

Memory System (4 Layers)

Layer 1: Daily Logs        → memory/daily/YYYY-MM-DD.md (raw truth, append-only)
Layer 2: Working Context   → chat history (ephemeral, in-session only)
Layer 3: Rolling State     → SUMMARY.md + LOOPS.md per agent/project
Layer 4: Curated Memory    → MEMORY.md (global + per-project)

Data flows up, never down:

You chat → daily log → PROPOSALS.md → coordinator reviews → MEMORY.md

Sub-agents can only write to PROPOSALS.md. Only the coordinator promotes to MEMORY.md. This prevents memory pollution.

Agent Roles (5 included)

Each agent has:

• Clear role definition
• Explicit "must NOT" constraints (prevents role bleed)
• Output format specification
• Tool access rules

Quick spawn (simple tasks)

sessions_spawn({
  task: "Build the user auth API",
  mode: "run",
  label: "backend-dev-auth"
})

Proper multi-agent (recommended)

# Create dedicated agents with their own workspace + identity
openclaw agents add backend-dev
openclaw agents add qa

# Each agent gets its own SOUL.md, AGENTS.md, tools, sessions
# Then spawn with auto-loaded context:
sessions_spawn({ task: "Fix the bug", agentId: "backend-dev", mode: "run" })

See memory/agents-spec.md for full specs, setup instructions, and role definitions.

Distillation Process

When to distill:

• Working → SUMMARY/LOOPS: every ~10-20 turns or when context feels large
• SUMMARY → MEMORY: milestones only (task done, decision made, lesson learned)
• Weekly: prune stale proposals, archive closed loops, dedupe memory

What never goes into long-term memory:

• Secrets, tokens, passwords
• One-off chat banter
• Unverified assumptions
• Transient debugging logs

See memory/distillation-process.md for the full process.

Adding a New Project

# Copy the template structure
cp -r memory/projects/_example memory/projects/my-project

# Edit the project memory
nano memory/projects/my-project/MEMORY.md

Automation (Optional)

Add these cron jobs for automated memory management:

Daily distillation (10pm) — auto-extracts facts from today's conversations into PROPOSALS.md:

Schedule: 0 22 * * * (your timezone)
Prompt: "Read today's daily log. Extract key decisions and lessons into PROPOSALS.md."

Weekly hygiene (Monday 10am) — reviews proposals, promotes accepted ones, prunes stale:

Schedule: 0 10 * * 1 (your timezone)
Prompt: "Review PROPOSALS.md files. Accept/reject. Promote accepted to MEMORY.md. Archive closed LOOPS."

Customization Guide

Minimal setup (single agent, no sub-agents)

Just use:

• SOUL.md — your AI's personality
• USER.md — about you
• MEMORY.md — long-term memory
• memory/daily/ — daily logs

That's it. The 4-layer system works even with just these files.

Full setup (multi-agent orchestration)

Use everything:

• All 4 layers
• Agent specs in memory/agents-spec.md
• Per-project memory folders
• PROPOSALS.md gate for sub-agent writes
• Automated distillation cron jobs

Credits

Built from real-world usage running a multi-agent AI workforce with OpenClaw.

Based on patterns from: Claude Code memory system, Mem0, Letta/MemGPT, CrewAI memory, and months of daily usage.

License

MIT — use it, fork it, improve it.