Basic Memory vs Supermemory
>Two products at the same $19 entry price, built on different ideas. A knowledge base people and agents write together, versus a memory engine that extracts facts from whatever you send it.
The short answer
Basic Memory is a knowledge base: notes are plain Markdown files that you, your team, and your AI tools read and write over MCP, with a web editor, comments, suggestions, file history, and sharing. Supermemory is a memory and context engine for developers, with a consumer app on top: you send it conversations, documents, or connected sources, and an LLM extracts facts and user profiles that your app searches. If you want memory people can read, edit, and own, at a flat price with a spending limit, use Basic Memory. If you want facts extracted automatically from many sources, behind an SDK, Supermemory is built for that.
| Basic Memory | Supermemory | |
|---|---|---|
| What it is | Knowledge base for people and AI agents | Memory and context API, plus a consumer app |
| Source of truth | Plain Markdown files you can open in any editor | Memories in a database; SMFS shows them as files |
| Who writes memory | People and agents, explicitly | An LLM extracts facts from what you send |
| Runs fully local without an LLM | Yes. SQLite and local embeddings | No. Extraction needs an LLM |
| People can edit it together | Yes. Web editor, comments, suggestions, live updates | No. App for saving and chatting with memories |
| File history and snapshots | Yes | No |
| Public sharing | Yes. Shared notes | No |
| Teams and roles | Roles, invite-only projects, audit log | Seats by plan; SSO on Enterprise |
| Connectors | No. Imports from Claude, ChatGPT, PDF, Word, PowerPoint, CSV | Yes. Google Drive, Gmail, Notion, GitHub, OneDrive, S3, web crawler |
| SDKs | CLI and REST API | Python and TypeScript |
| MCP and coding agents | Yes. Claude Code, Codex, Cursor and any MCP client | Yes. Claude Code, Cursor and others |
| Open source | Yes (AGPL-3.0) | Yes (MIT) |
| Entry price | Cloud $19/month: 1,900 credits, 2 members | Pro $19/month: $20 of credit, 3 seats |
| How usage is charged | Flat plan with included credits and a spending limit | Per token written ($5 per million for text), plus searches |
| Team price | Business $249/month, up to 25 members | Scale $399/month, unlimited seats |
Written by people and agents vs extracted by an LLM
Supermemory is built around automatic memory. You send it chat transcripts, documents, PDFs, images, or video, or connect Gmail, Google Drive, and Notion, and an LLM pulls out facts and keeps a profile of each user. Your agent then searches those memories. For an app that needs memory to appear without anyone deciding what to save, that is a real strength, and its connectors bring in sources Basic Memory does not.
Basic Memory keeps the knowledge itself. Notes are written on purpose, by you or by your AI as you work, and stay as full Markdown with links between them. Nothing is summarized away, and you can read exactly what your AI knows. Basic Memory agents can also curate a project on instructions you write in a note, so maintenance does not have to be manual.
Files you can open vs files as a view
Supermemory’s SMFS presents memories to coding agents as a file system, which shows how useful files are for agents. In Basic Memory the files are the source: every note is a Markdown file you can open in Obsidian or VS Code, version with git, or copy out. Agents read them with familiar tools such as ls, cat, and grep.
People get the same files in a web app built for working together: an editor, comments and suggestions, file history and snapshots, and notes you can share publicly. Supermemory’s app is for saving and chatting with memories, not for editing them together.
Flat price vs per-token pricing
Both start at $19 a month. Supermemory’s plans include usage credit, then charge per token written ($5 per million tokens of text, $10 for rich content) and per search, because each write runs an LLM. An agent that saves 2 million tokens of transcripts a month spends about $10 to $14 on writes alone.
Basic Memory charges nothing per token written. Each plan includes credits for compute and storage, and the Owner sets a spending limit, so the monthly bill is predictable. For teams, Business is $249 a month for up to 25 members; Supermemory’s team plan, Scale, is $399 a month.
Choose Basic Memory if
- You want memory people and agents both read and write, in plain Markdown you own
- You want a shared workspace: editor, comments, suggestions, file history, sharing
- You want a flat monthly price with a spending limit, not per-token charges
- You want to run it free on your own machine with no LLM bill
Choose Supermemory if
- You want facts extracted automatically from conversations and documents
- You need connectors that pull in Gmail, Google Drive, Notion, or GitHub
- You are building an app and want Python or TypeScript SDKs
Frequently asked questions
Is Basic Memory a Supermemory alternative?
Yes, when you want memory people can read and edit. Both give AI tools long-term memory over MCP and start at $19 a month. Basic Memory stores knowledge as Markdown files with a collaborative web app and a flat price; Supermemory extracts facts with an LLM and charges per token written.
How does pricing compare?
Basic Memory Cloud is $19 a month with 1,900 credits and 2 members, Pro is $99 with 5 members, and Business is $249 with up to 25 members, each with a spending limit and no per-token charges. Supermemory Pro is $19 with $20 of credit and 3 seats, Max is $100, and Scale is $399 with unlimited seats, with usage charged per token written and per search. Prices checked October 2026.
Does Basic Memory extract facts automatically?
Not on every write. Your AI writes notes as you work, and Basic Memory agents can curate a project on instructions you write in a note. Supermemory runs LLM extraction on everything you send it, which is more automatic and charged per token.
Are both open source?
Yes. Basic Memory is AGPL-3.0 at github.com/basicmachines-co/basic-memory, and Supermemory is MIT. Basic Memory runs fully local with SQLite and no LLM key; Supermemory’s local version needs an LLM for extraction.
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.