supermemory
Supermemory is the 'second brain as an API' approach. It auto-ingests from your browser, Google Docs, Slack, Notion, and other sources — then exposes the consolidated brain as an API your agent calls. The fastest way for a solo founder to stand up a personal brain that already has all their context in it.
- Solo founders with content scattered across 10 SaaS apps
- Anyone who wants zero-touch ingestion — no manual note-taking
- Teams that want a managed option (hosted) or self-host (OSS)
What you'll do
Supermemory is both an OSS project and a hosted service. For the OSS path, you install the TS SDK and run the server locally. Budget 15 minutes.
Before you start
- Node.js 20+
- An OpenAI API key (or use supermemory.ai managed, which handles this)
Step-by-step install
- 011. Install the SDK
npm install supermemory
- 022. Set your API key
Either get one from supermemory.ai (managed) or set OPENAI_API_KEY and self-host the open-source server from the repo.
- 033. Add a source
Use the SDK to connect your browser history, Slack, Google Docs, or Notion. Each source has a 3-click setup flow.
import Supermemory from "supermemory" const sm = new Supermemory({ apiKey: "..." }) // Add a source await sm.sources.addBrowserHistory({ userId: "vedant" }) - 044. Query from your agent
Your agent calls the supermemory API for context retrieval.
const context = await sm.search({ userId: "vedant", query: "What did I read about AI memory tools this week?" }) - 055. Optional: add the MCP bridge
Supermemory has an MCP server you can point Claude Code or Cursor at — same context, exposed as an MCP tool.
Your first 10 minutes
- 01Wire your browser history. Search it for 'AI memory tools this week'. Should recall real pages you visited.
- 02Add Slack as a second source. Search 'what did Amara and I discuss about Q3'.
- 03Review the ingested memory — supermemory has inspection tools. Prune anything noisy.
- 04Set ingestion cadence appropriately (daily is usually enough for browser).
- 05Add Cognition CLO for retention modeling across your team.
Troubleshooting
Ingestion is noisy — results include junk pages.
Tune the ingest filter. Supermemory supports URL-pattern excludes — use them aggressively.
The managed service is too expensive.
Self-host the OSS server. The repo has a one-file Docker setup for a personal instance.
supermemory holds the knowledge. Cognition CLO models retention per employee per concept using a Weibull forgetting curve — so you see decay before it becomes a missed SOP or a failed audit.