Install
openclaw skills install obsidian-memory-systemStructured persistent memory system using an Obsidian vault with daily journals, project docs, knowledge base, self-improvement logging, and Discord workspace integration. Use when: (1) Setting up a new OpenClaw agent's memory system, (2) Agent needs persistent memory across sessions, (3) Organizing project documentation, daily journals, or knowledge base, (4) Logging errors, learnings, or feature requests for continuous improvement, (5) Setting up Discord as primary workspace (voice, components, threads, channel architecture), (6) User says 'set up memory', 'initialize vault', 'create journal', 'log this error', 'remember this', 'update memory', 'set up discord', or 'migrate from whatsapp'. Also covers semantic search setup, promotion pipeline for learnings into brain files, and Discord workspace configuration.
openclaw skills install obsidian-memory-systemPersistent agent memory using an Obsidian vault with structured folders, daily journals, semantic search, and self-improvement logging.
~/clawd/ ← OpenClaw workspace
├── SOUL.md ──symlink──→ vault/00-brain/SOUL.md
├── USER.md ──symlink──→ vault/00-brain/USER.md
├── AGENTS.md ──symlink──→ vault/00-brain/AGENTS.md
├── TOOLS.md ──symlink──→ vault/00-brain/TOOLS.md
├── MEMORY.md (copy, NOT symlink — indexer skips symlinks)
├── HEARTBEAT.md (standalone, periodic tasks)
├── memory/ (real dir with copies — synced from vault/10-journal/)
├── scripts/sync-memory.sh (rsync vault→memory every 30 min via cron)
└── vault/ ← Obsidian vault
├── 00-brain/ Core identity files
├── 10-journal/ Daily work logs (YYYY-MM-DD.md)
├── 20-projects/ Project docs (overview, decisions, timeline)
├── 30-knowledge/ Reusable reference material
├── 40-people/ People notes
├── 50-ideas/ Future plans, brainstorms
├── 60-learnings/ Self-improvement logs (errors, learnings, feature requests)
└── templates/ Note templates
OpenClaw auto-loads workspace root files (SOUL, USER, AGENTS, TOOLS, MEMORY) every session. Symlinks bridge workspace ↔ vault so Obsidian and the agent see the same files.
⚠️ Memory indexer symlink limitation: OpenClaw's memory indexer (memorySearch) uses lstat and explicitly skips all symlinks — both directories and files. Use real file copies for MEMORY.md and memory/ with a sync script (see references/discord-setup.md → Memory Integration).
Run the setup script to initialize everything:
bash scripts/setup-vault.sh ~/clawd
This creates the vault structure, brain file templates, symlinks, vault/60-learnings/ directory, and templates. Then customize each brain file for your agent.
For details on each file's purpose: read references/brain-files.md
For OpenClaw config: read references/openclaw-config.md
vault/10-journal/YYYY-MM-DD.mdmemory_search for any recall needsmemory_search("query") to find past context[[20-projects/name/overview|Name]]vault/10-journal/YYYY-MM-DD.md (always)vault/20-projects/*/overview.md (if project changed)| Situation | Action |
|---|---|
| Command/operation fails | Append to vault/60-learnings/ERRORS.md |
| User corrects you | Append to vault/60-learnings/LEARNINGS.md (category: correction) |
| Found better approach | Append to vault/60-learnings/LEARNINGS.md (category: best_practice) |
| Knowledge was outdated | Append to vault/60-learnings/LEARNINGS.md (category: knowledge_gap) |
| User wants missing feature | Append to vault/60-learnings/FEATURE_REQUESTS.md |
Entry format — see references/logging-format.md
When a learning proves broadly applicable, promote it:
| Learning Type | Promote To | Then set Status → promoted |
|---|---|---|
| Behavioral patterns | SOUL.md | |
| Workflow improvements | AGENTS.md | |
| Tool gotchas | TOOLS.md | |
| Cross-project decisions | MEMORY.md |
Promote when: Recurrence ≥ 3, seen across 2+ tasks, within 30-day window.
Always connect related content:
See [[20-projects/my-app/overview|My App]] for details.
Built by [[40-people/juan|Juan]].
Every vault markdown file needs:
---
title: Document Title
type: note|project|reference|daily|decision
created: YYYY-MM-DD
permalink: agent-name/folder/filename
---
read references/memory-rules.mdDuring periodic heartbeats (every few days):
vault/60-learnings/ entriesDiscord replaces WhatsApp as primary channel with major upgrades: streaming, voice, buttons, threads, and channel isolation.
For full setup guide: read references/discord-setup.md
| Feature | What it does |
|---|---|
| Channel separation | Per-project channels = isolated sessions, less token waste |
| Streaming | See responses as they generate |
| Voice channels | Real-time voice: Whisper STT → LLM → OpenAI TTS |
| Interactive components | Buttons, selects, forms for quick decisions |
| Thread-bound coding | Codex/Claude Code get their own threads |
| Reactions | Visual ack (configurable emoji) |
| Auto-presence | Bot status shows health |
🏠 Home → #general, #tasks, #coding
🔊 Voice → 🎙 General (voice conversations)
🏥 [Your Org] → #project-a, #project-b, ...
🤖 Agents → #agents (thread-bound sessions)
📋 Ops → #logs, #cron
🧪 Research → #research
Set channel topics with vault pointers for project routing:
Project A app | vault: 20-projects/project-a/ | repo: user/project-a | port: 3001
{
"channels": {
"discord": {
"enabled": true,
"token": "YOUR_BOT_TOKEN",
"groupPolicy": "allowlist",
"dmPolicy": "allowlist",
"allowFrom": ["YOUR_USER_ID"],
"guilds": {
"YOUR_GUILD_ID": {
"requireMention": false,
"users": ["YOUR_USER_ID"]
}
},
"streaming": "partial",
"replyToMode": "first",
"historyLimit": 30,
"threadBindings": { "enabled": true, "spawnSubagentSessions": true, "spawnAcpSessions": true },
"ackReaction": "🦅",
"autoPresence": { "enabled": true, "healthyText": "Online" }
}
},
"tools": {
"profile": "full",
"exec": { "security": "full", "ask": "off" }
},
"messages": { "ackReactionScope": "all" }
}
⚠️ Without the guilds block, the bot only works in DMs. This is the #1 setup issue.
Full production config with status reactions, custom emoji, voice, components, and troubleshooting: read references/discord-setup.md
Full OpenClaw config reference: read references/openclaw-config.md
# Count pending learnings
grep -rh "Status\*\*: pending" vault/60-learnings/*.md | wc -l
# Find high-priority items
grep -B5 "Priority\*\*: high" vault/60-learnings/*.md | grep "^## \["
# Search learnings by area
grep -l "Area\*\*: backend" vault/60-learnings/*.md