Obsidian memory for openclaw?

[left cradle]

Any chance we can install a proper setup obsidian vault for openclaw ? any official guides or steps?

[final frost]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[tropic hinge]

I can help you with that if you need it

[mystic moth]

Yes — please do share your working recipe here too. The “official” path depends on what they want:

1) Use an existing Obsidian vault as searchable memory:
Use QMD and add the vault as an indexed path via memory.qmd.paths. Good for “let OpenClaw search my notes.” (docs.openclaw.ai)
Docs: https://docs.openclaw.ai/concepts/memory-qmd

2) Let OpenClaw maintain an Obsidian-friendly knowledge vault:
Use the bundled memory-wiki plugin with vault.renderMode: "obsidian". Good for structured pages, claims, provenance, dashboards, etc. (docs.openclaw.ai)
Docs: https://docs.openclaw.ai/plugins/memory-wiki

Minimal shape:

{
  memory: {
    backend: "qmd",

    qmd: {
      paths: [
        { name: "obsidian", path: "~/Documents/Obsidian/MyVault", pattern: "**/*.md" }
      ]
    }
  },
  plugins: {
    entries: {
      "memory-wiki": {
        enabled: true,
        config: {
          vaultMode: "isolated",
          vault: {
            path: "~/Documents/Obsidian/OpenClaw-Wiki",
            renderMode: "obsidian"

          }
        }
      }
    }
  }
}

Then check with:

openclaw memory status --deep
openclaw memory index --force
openclaw wiki doctor
openclaw wiki init

If you want exact steps for their machine, first get openclaw --version, OS, and where their Obsidian vault lives.

[arctic marsh]

@mystic moth Obsidian has the free plan, sync plan and publish plan. The publish plan has the features "Graph and full text search" does graph and full text search help openclaw?

[mystic moth]

Short answer: no, don’t buy Obsidian Publish just for OpenClaw.

Publish’s graph/search are mainly for humans browsing a published website. OpenClaw works better from the local Markdown vault: it can use its own memory_search, QMD indexing, or memory-wiki/Obsidian-friendly mode. (obsidian.md)

Best setup:

• Free Obsidian: enough if the vault is on the same machine as the OpenClaw Gateway.
• Obsidian Sync: useful if you need the vault synced onto the Gateway machine. (help.obsidian.md)
• Obsidian Publish: useful if you want a public/private website for people, not for OpenClaw memory. (help.obsidian.md)

For OpenClaw, the useful part is pointing memory/search at the vault, e.g. via memorySearch.extraPaths or QMD memory.qmd.paths, then OpenClaw indexes the Markdown directly. The graph visualization itself doesn’t really add anything for the agent.

[arctic marsh]

I have agents on a mac mini and agents on a VPS. It would be good if they all share the same memory/information so I could get sync to help with that

[mystic moth]

Yes — Obsidian Sync can be useful here, but as the file sync layer, not because of Publish/Graph.

Best pattern:

1. Create one Obsidian vault, e.g. OpenClaw Knowledge.
2. Sync that vault to both the Mac mini and VPS. Obsidian now has Headless Sync for command-line/server syncing without the desktop app. (obsidian.md)
3. Configure each OpenClaw install to index that vault.

Simple config option:

{
  agents: {
    defaults: {
      memorySearch: {
        extraPaths: ["/path/to/OpenClaw Knowledge"]

      }
    }
  }
}

Then run:

openclaw memory index --force
openclaw memory status --deep

Safer recommendation: let each agent keep its own normal MEMORY.md, and use the synced Obsidian vault as a shared knowledge base. If multiple agents write to the same synced files at once, you can get conflicts. A good structure is:

OpenClaw Knowledge/
  Shared/
  Projects/
  People/
  Decisions/
  Agent Inbox/
    mac-mini/
    vps/

So: Sync plan yes, Publish plan no. OpenClaw wants the local Markdown files; the Graph/full-text Publish features are mostly for humans browsing a site.

OpenClaw docs worth reading:
https://docs.openclaw.ai/concepts/memory
https://docs.openclaw.ai/concepts/memory-qmd
https://docs.openclaw.ai/reference/memory-config

[arctic marsh]

Should each agent have their own / folder? So Frigg and Freya could store the agent files in here? Or should only their individual memories go there from /.openclaw/workspace-frigg/memory/ and /openclaw/workspace-freya/memory/ ?

[mystic moth]

I’d split it into three layers:

1. Each agent gets its own OpenClaw workspace

   • ~/.openclaw/workspace-frigg/
   • ~/.openclaw/workspace-freya/

   This is where their AGENTS.md, SOUL.md, USER.md, MEMORY.md, memory/YYYY-MM-DD.md, etc. live. Keep these separate because they’re each agent’s identity + private operating memory.

2. One shared Obsidian vault

   • Example: ~/Obsidian/OpenClaw Knowledge/
   • Sync this to Mac mini + VPS.
   • Configure both agents to index it via memorySearch.extraPaths.

3. Per-agent inbox folders inside the vault

   OpenClaw Knowledge/
   

     Shared/
     Projects/
     People/
     Decisions/
     Agent Inbox/
       frigg/
       freya/

So I would not make the synced Obsidian vault replace workspace-frigg / workspace-freya. Let OpenClaw own those workspaces. Use Obsidian as the shared knowledge base that both agents can search and optionally write curated notes into.

Config shape:

{

  agents: {
    list: [
      {
        id: "frigg",
        workspace: "~/.openclaw/workspace-frigg",
        memorySearch: {
          extraPaths: ["~/Obsidian/OpenClaw Knowledge"]
        }
      },
      {
        id: "freya",
        workspace: "~/.openclaw/workspace-freya",
        memorySearch: {
          extraPaths: ["~/Obsidian/OpenClaw Knowledge"]
        }