Supermemory alternatives: Substrate vs Supermemory, where each wins

Substrate and Supermemory are two different products from two different companies, despite the similar-sounding names. Supermemory is memory infrastructure for developers who build agents or applications on top of it: an API, plugins, an MCP server and connectors. Substrate is a memory that you and your agents use directly: you can read and edit the same memory that your agents recall. If you are putting memory into a product for your own users, Supermemory is the closer fit. If you run a small team with a fleet of agents and want them to share one context, Substrate is.

Substrate (trysubstrate.co) wrote this comparison, so it is not neutral. We try to say plainly where Supermemory is better. The Supermemory facts below come from its public docs, pricing page and GitHub repository, as of November 5, 2026. Check them again before you decide, because both products change.

Is Substrate the same as Supermemory?

No, we are different products from different companies.

Supermemory refers to itself in its documentation as "Context infrastructure for AI agents". Developers can use its API to implement memory, retrieval and user profiles in their own agents and products.

Substrate is a hosted product: you sign up, connect your agents and your sources, and then you can read and edit the same memory that your agents recall. You can also share a page with other people by email invite, with view-only access, and you can revoke access at any time.

Also, there are other companies called Substrate, so please check that you are on the right domain: ours is trysubstrate.co.

Supermemory alternatives: which one fits?

The answer depends on the job you want to do.

• Building memory into your own product, for your own users: stay with Supermemory, or compare other developer memory libraries such as Mem0. We wrote a separate comparison: Mem0 alternatives: Substrate vs Mem0.
• One person on one agent, such as Claude Code: the memory built into the tool may be enough, and you may not need anything extra. We cover that in Claude Code memory: making it remember across sessions.
• A small team with a fleet of agents (Claude Code, Codex, Cursor, ChatGPT) that should share context, where the humans also need to read it: that is what Substrate is for. See how to share context between Claude and Cursor for the manual version.
• Agents that live in one tool, with nobody else needing the context: skip all of this.

If you want to understand the category first, MCP memory servers: what they are, how to set one up, and how to choose explains it.

Supermemory pricing vs Substrate pricing

Supermemory is cheaper to start, since it has a free tier, and its price grows with usage. Substrate is one flat price. As of November 5, 2026, the Supermemory pricing page lists:

• Free: $0 per month, with $5 of monthly credits and 1 seat
• Pro: $19 per month, with $20 of credits and 3 seats
• Max: $100 per month, with $130 of credits and a Gmail connector
• Scale: $399 per month, with $600 of credits, SOC 2, a HIPAA BAA and a self-hosted option
• Enterprise: custom pricing

On top of that, usage is metered and taken from your credits. For example, plain text memory is $0.005 per 1K SM tokens (rich content $0.010) and search is $0.005 per 1K queries. So you pay more if you use the product more, which can work well for a product with many end users.

Substrate is $20 per month on one plan, for one person and all of their agents, with a 7-day trial and no card needed.

Supermemory MCP vs Substrate MCP

Both connect to agents over MCP, and both let a person decide what is connected. Supermemory runs a hosted MCP server at https://mcp.supermemory.ai/mcp. You sign in with OAuth in the browser, you don't need an API key, and you choose which spaces the client can access. In its MCP docs the tools include search_memory, add_memory, forget_memory, get_profile and list_spaces. A space isolates one body of work, and teammates can collaborate in the spaces they are permitted to read or write. You can change a connected app's access, or disconnect it, on the Agents page of the Supermemory console.

Substrate also connects agents over MCP: claude.ai, ChatGPT (as a custom connector on the web, on plans that allow it), Cursor, OpenClaw and Grok CLI. Claude Code, Cowork, Codex and Hermes connect through a plugin instead. In Substrate a human approves and revokes each agent connection.

The difference is what sits behind the connection. In Supermemory it is a developer-facing memory API with spaces. In Substrate it is a memory that you can read and edit yourself.

Does Supermemory work with Claude Code?

Yes, through a plugin. Add the marketplace, then install it:

/plugin marketplace add supermemoryai/claude-supermemory
/plugin install supermemory

You need Node.js 18 or newer and a Supermemory API key. A browser login saves the key for you. Once it runs, three things happen:

1. At the start of a session, it loads profile facts for the current project.
2. Before each substantive prompt, it searches the memory and injects relevant matches.
3. After each response, it saves the new conversation content in the background.

The plugin saves to a repo container that is shared with Supermemory's Codex, OpenCode and Cursor plugins. These details come from Supermemory's Claude Code docs, as of November 5, 2026.

Substrate also has a Claude Code plugin. The same Substrate memory can be reached from Codex (plugin) and from Cursor (MCP), and from ChatGPT as a custom MCP connector where your plan allows it. Both give Claude Code memory across sessions.

Is Supermemory open source?

Mostly, yes. The supermemoryai/supermemory repository on GitHub is MIT licensed, and its README says the plugins and the MCP server are open source. You can also run it locally with npx supermemory local, which serves the API on localhost:6767.

Supermemory's own docs don't name a license, so the repository is the best place to check.

Substrate is a hosted product, so there is no self-hosted version. If you need to run memory on your own machine or inside your own network, that is a clear reason to pick Supermemory.

Substrate vs Supermemory: where does each one win?

Supermemory wins for builders. Substrate wins for a person or small team running several agents.

Choose Supermemory if:

• You are a developer building memory into a product that many end users will use.
• You need open source or self-hosting.
• You want an API with usage-based pricing and a free tier.
• You need compliance such as SOC 2 or HIPAA, which its pricing page lists on the Scale plan.
• You want connectors such as Google Drive, Gmail, Notion, OneDrive, GitHub, Granola or a web crawler.

Choose Substrate if:

• You run several agents (Claude Code, Codex, Cursor, ChatGPT) and want them to share one memory.
• You want to read and correct what the agents recall.
• You keep context in Gmail, Notion (both read-only), WhatsApp (where enabled) or uploaded files, and want it in a memory you can read and correct.
• You want to share a page with a named person, view-only.
• You want to approve and revoke agent connections yourself.
• You prefer one flat price of $20 per month.

You can also use both: Supermemory as infrastructure for your product, and Substrate as the memory for you and your agents.

Substrate's own benchmark claim is "State-of-the-art at finding and connecting insights: 66%+ on KnowMe-Bench." We have not tested Supermemory against it, so it says nothing about how the two compare.

Substrate: AI memory for your team, editable, permissioned, owned. You can try it free for 7 days, no card needed, at trysubstrate.co.