Basic Memory vs Obsidian
>Not really either-or. Obsidian is where you write and browse your notes. Basic Memory is how your AI finds, reads, and writes them, on your machine or in the cloud.
The short answer
Keep your vault. Point the open-source Basic Memory at it (bm project add my-vault --local-path ~/path/to/vault) and it indexes every note, links, tags, and frontmatter included, so your AI searches the vault instead of grepping it, and writes notes Obsidian shows the moment they land. That is free and stays on your machine. Basic Memory Cloud is for when one machine is not enough: the same memory on every device, in Claude, ChatGPT, Codex, and any MCP client, shared with your team, with agents that keep it organized. Obsidian stays the best place to write and browse; Basic Memory is the memory your AI tools share.
| Basic Memory | Obsidian | |
|---|---|---|
| What it is | Memory your AI tools read and write, over MCP | A Markdown note-taking app for people |
| Source of truth | Plain Markdown files | Plain Markdown files (the same ones) |
| How AI finds a note | Full-text, semantic, and knowledge-graph search, built in | Not built in. AI tools read the files or rely on community plugins |
| AI reads | The notes a search returns, or one section of a note | Whole files |
| AI writes | Notes, edits, and links, with checks against overwriting newer edits | Not built in |
| Works on an existing vault | Yes. Point a project at the folder; no plugin needed | — |
| Runs fully local, free | Yes. Open source (AGPL-3.0), SQLite, local embeddings | Yes. The app is free |
| Same memory in Claude and ChatGPT | Yes, with Basic Memory Cloud | No |
| Sync across devices | Basic Memory Cloud, plus a web app on any device | Obsidian Sync, $4 per user per month (annual), end-to-end encrypted |
| An editor your AI works in too | Yes. Your AI’s edits, comments, and suggestions appear in real time; you accept or reject them | No. AI changes land as file edits |
| Browser agents (WebMCP) | Yes. Agents in the browser work in the app as you | No |
| Share with a team | Members, roles, invite-only projects, comments, suggestions | Shared vaults through Obsidian Sync |
| Agents that keep notes current | Yes. Basic Memory agents, reviewed and undoable | No |
| Graph view, themes, plugins | A web app with a wiki for each project; use Obsidian for the rest | Yes. Thousands of community plugins |
| Price | Free locally. Cloud from $19/month | Free. Sync and Publish are paid add-ons |
A vault is the right format. AI just cannot query it
Markdown with wiki links and frontmatter is exactly the shape long-term memory should have, and Obsidian users already have years of it. The trouble starts when an AI tool works on that folder directly: it greps for words, opens whole files to check them, and starts over the next session. On a large vault that is slow, misses notes that use different words, and fills the context window with text it did not need.
Basic Memory keeps the files and adds what the AI needs on top: full-text and semantic search, so it finds the note about “pricing” when you wrote “plans”; a knowledge graph built from your links, so it can follow a decision to the meeting it came from; and reads of one section instead of the whole file. Your AI gets to the right note in one step and reads only what it uses.
Local and free: point it at your vault
The open-source Basic Memory drops into an existing vault. Add a project at the vault’s folder and it starts indexing everything, then watches for changes. Notes your AI writes appear in Obsidian immediately, with clickable backlinks and a place in the graph view; notes you edit in Obsidian are what your AI reads next. No plugin, no import, and the files stay ordinary Markdown.
This runs entirely on your machine with SQLite and local embeddings, with no account and no API key. If you want AI memory for one computer and one set of AI tools, that is what it is for.
Basic Memory Cloud: memory every tool and device shares
Basic Memory Cloud is for when memory has to travel. The same knowledge is in Claude on your phone, ChatGPT on the web, Claude Code and Codex in your terminal, and any other MCP client, so what one AI learns, the others can use. A web app opens it on any device, and you can keep editing in Obsidian and sync your vault with the CLI.
It is also where memory becomes shared: invite your team with roles and invite-only projects, comment and suggest edits in the note, share a note publicly, and let Basic Memory agents file, link, and summarize on a schedule, with every change on the record. Obsidian Sync syncs files between your devices; Basic Memory Cloud gives your AI tools and your team one memory.
An editor that works for you and your AI
In Obsidian, an AI that changes your notes is editing files behind the app. In the Basic Memory Cloud editor, your AI works in the same note you have open. When it updates the note, the note updates on your screen in real time, without reloading. It can leave a comment, or suggest an edit inline for you to accept or reject, so you review its work where you read it. If you and your AI change the same note at once, a merge view shows both versions and you choose.
In browsers that support WebMCP, agents that live in the browser get tools in the app itself: they search and open notes, suggest edits, and check live notes, acting as you. Your other AI tools keep working over MCP, on the same notes.
Choose Basic Memory if
- You want your AI to search, read, and write your vault, not grep it
- You want it free and local: point the open-source Basic Memory at your vault
- You want the same memory in Claude, ChatGPT, Codex, and on every device (Cloud)
- You want an editor where your AI comments, suggests, and edits beside you, in real time (Cloud)
- You want to share that memory with a team, or have agents keep it organized (Cloud)
Obsidian alone is enough if
- You write and browse notes yourself and do not need AI to use them
- You need end-to-end encrypted sync across your devices
- Your workflow depends on Obsidian plugins, Canvas, or Publish (keep them: Basic Memory works on the same files)
Frequently asked questions
Does Basic Memory replace Obsidian?
No. They work on the same Markdown files. Obsidian is where you write and browse; Basic Memory gives your AI tools search, a knowledge graph, and write access to those files. Most people who use Obsidian keep it and add Basic Memory.
How do I use Basic Memory with an existing Obsidian vault?
Install the open-source Basic Memory and add a project at your vault’s folder: bm project add my-vault --local-path ~/path/to/vault. It indexes every note and watches for changes, and notes your AI writes show up in Obsidian right away. See the Obsidian guide.
Why not just give my AI the vault folder?
It works on small vaults, but an AI with only a folder can grep for words and open whole files. It misses notes that use different words, cannot follow your links, and spends its context window reading files to find one fact. Basic Memory search returns the right notes and their connections, and the AI can read one section instead of the whole file.
When should I use Basic Memory Cloud instead of local?
When you want the same memory on more than one device or in more than one AI tool, such as Claude on your phone and ChatGPT on the web, or want to share it with a team or run Basic Memory agents. Local is free and stays on one machine. Cloud starts at $19 a month.
Can my AI work in the editor with me?
In Basic Memory Cloud, yes. When your AI updates a note, the note you have open updates in real time, and it can comment or suggest edits for you to accept or reject. In browsers that support WebMCP, browser agents work in the app as you. Obsidian has no built-in way for an AI to comment or suggest; AI changes arrive as edits to the files.
Is my vault still mine?
Yes. Notes stay plain Markdown files. Locally they never leave your machine; in Basic Memory Cloud you can export or sync them back down at any time.
Start building with your AI.
Try it for 7 days. Your notes stay yours either way. Start building a knowledge base that grows with you.