MegaMem Best Practices
13 practices for getting the most out of MegaMem — for all skill levels
Abstract
MegaMem is not a note organizer with AI features bolted on. It’s a bridge between the old world and the new — between traditional docs and spreadsheets and the knowledge graph engine we like to call the Synthetic Cortex. And like any infrastructure, it rewards the people who learn how to use it well. These are the best practices that make the difference between knowing the path,
and walking the path.
You installed MegaMem. Your vault is connected to a graph database. The sync button works. Now what? Why isn’t it living my life for me?
Do I really have to read all the docs and learn Obsidian?!?
No. But you do have to start thinking differently.
Before I built MegaMem I dreamt in spreadsheets. Now I dream in interconnected nodes and relationships. The tool isn’t the problem. The mental model is. Here are some best practices and shortcuts that can get you there faster.
The Mental Model: Vault, Graph, Session
MegaMem operates across three layers. Confusing them is the single biggest source of frustration for new users.
The Vault is your document and task management interface — like Google Docs + Notion + Asana rolled into one. Obsidian is the user interface, but you almost never touch it.
The Graph is the machine-readable layer: a Synthetic Cortex of entities extracted from your docs and sessions, linked by relationship edges, timestamped as they change over time.
The Session is ephemeral — a single chat thread in your LLM client, what the current conversation knows. This is what MegaMem MCP plugins into, but when not used correctly — all the juicy intelligence that you generate dissolves when the session ends (like it does in everywhere else).
A note exists in the Vault. A synced note with meaningful content exists in both the Vault and the Graph. A conversation insight only lives in your Session until you capture it — and most people never do.
See all those items in your recent chat history? Your LLM picked a couple of facts from each and stored them in “Memory” — the rest is buried and gone.
FYI: You could stash all those chats in your vault with a single MegaMem command: “At the start of every chat, create a ChatHistory note in mv, and update it as we go. Capture concise summaries from you as ‘Agent:’, and full responses from me.”
But why would you? That’s your old brain running old code. Upgrading your LLM tools will only get you so far — it’s time to upgrade your internal OS.
The Fastest Shortcut: One System Prompt
Before diving into individual practices, here’s the single thing you can do right now that encapsulates most of the items below. Add this to Claude Desktop → Settings → Personal Preferences and it applies to every conversation automatically:
(For non-MegaMemPro users, adjust accordingly)
# MegaMem Tools
MegaMem MCP tool schemas are provided directly in this system prompt — use them immediately without calling tool_search first.
**"Create a note":** Use `megamem:create_note_with_template` with template `TPL Note`, place in `01_Inbox`. Add 1–2 relevant tags to frontmatter (no # prefix — Obsidian tags in frontmatter are plain text, e.g. `tags: [reminder, dev]`).
**"Create a doc in mv":** Use `megamem:create_note_with_template` — choose the most appropriate template and folder based on context. If the 'megamem-pro' Skill is installed, follow the workflow.
**"Create a task":** Use `megamem:create_note_with_template` with `request_type: TPL Task` and `file_name: <task title>`. By default (no `target_folder`), Templater routes to the personal tasks folder. When working within a specific project, set `target_folder` to that project's `05_Tasks` subfolder.
**"Search mm" / "Ask mm":** Use `megamem:search_memory_facts` or `megamem:search_memory_nodes` with the default group_id. Then (as necessary) use `megamem:get_episodes` and `megamem:get_entity_edge`.
You also have three skills installed: `megamem`, `megamem-pro`, and `megamem-skills`. When unsure which mv or mm tool to use, read `megamem` skill's `references/reference.md` for full parameter details. When creating notes, the `megamem-pro` skill has the TemplateRegistry workflow. When user says 'mms' or 'mmSkill' — their intention is to use skills stored in their vault, so use the 'megamem-skills' workflow.
**CRITICAL:** When creating ANY file in mv (notes, docs, tasks, records, specs — regardless of how the request is framed), ALWAYS use `megamem:create_note_with_template`. NEVER use `megamem:create_obsidian_note` for structured content. If unsure which template, query the TemplateRegistry first.
That’s it. Paste it once and your LLM knows how to use MegaMem correctly from that point forward.
The full Best Practices guide — the complete version you can feed directly to your LLM — is available in your MegaMemPro folder at MegaMemPro/Resources/Guides/MegaMem Guide - 03 Best Practices.
Best Practices For Everyone
1. Start with Bjørn’s Vault
The single highest-leverage thing a new user can do is start with Bjørn’s Vault rather than a blank Obsidian setup.
Almost all LLMs are pre-engineered to follow existing patterns. Give one a blank vault and it makes structural decisions based on nothing with confidence — and it is almost always “Strong and Wrong”. The result: folder sprawl, inconsistent structure, and a graph that slowly becomes harder to query meaningfully.
Bjørn’s Vault provides a proven, pre-engineered starting point: 10 pre-configured plugins, a Johnny.Decimal folder structure optimized for LLM-first workflows, templates with frontmatter auto-routing, pre-configured Bases, and the MegaMem Ontology — the most valuable piece for anyone planning to use the graph seriously. It can save technical users days of setup time and non-technical users months of trial-and-error.
Available at endogon.com/megamem
2. Always Use Templates
Every note should start from a template. This isn’t optional polish — it’s the foundation of a queryable vault.
Templates do three things: they enforce the type field (how MegaMem classifies notes as entity types in the graph), they auto-route notes to the right folder via Templater, and they give your LLM consistent structure to write into every time.
The mechanism behind this is called The Skeleton. When you call create_note_with_template, MegaMem creates the note and sends the full template skeleton back to your LLM in the same response — invisible to you. The LLM immediately populates it. Every template is effectively structured prompt logic: it tells the LLM what to write, in what format, with what properties. The templates in Bjørn’s Vault are a strong starting point, but the real power comes when you define templates for your own document types — including structured content areas and relational instructions.
3. Know What to Sync — and What Not To
