Supermemory sells context infrastructure — memory, search, user profiles, and connectors (Notion, Drive, Gmail, PDFs, audio). ReadmeDB sells a shared project brain: versioned markdown files plus optional distilled memories, wired to Cursor / Claude / Windsurf via MCP.
| Dimension | ReadmeDB | Supermemory |
|---|---|---|
| Primary model | Hosted markdown namespaces agents edit | Context cloud: ingest → embed → retrieve |
| Best job | Cross-tool coding brain (Cursor ↔ Claude) | Apps that need broad personal/org recall |
| Human-readable store | .md files + dashboard diffs | Chunks / memories in a cloud UI |
| Agents write durable truth | First-class (write/patch/version) | Add memories / ingest docs |
| Connectors (Drive/Gmail/Notion) | Not built | Core product surface |
| Coding-agent MCP | File + memory tools for IDEs | API / plugins — IDE-native varies |
| Auto-extract from chat | Extract endpoint (opt-in) | Memory pipeline focus |
| Pricing vibe | Simple free → Pro hosted brain | Free → $19–$399 context tiers |
| Typical buyer | Indie hackers & small eng teams | Builders needing multi-source RAG memory |
You need to suck in Gmail, Notion, Drive, and PDFs and retrieve across all of it for a consumer or enterprise agent. Connectors + broad ingestion is their lane — ReadmeDB is not trying to be your document crawler.
Your pain is agents re-asking stack/decisions every session, or Cursor and Claude not sharing one source of truth. You want CLAUDE.md / decisions.md as living files — humans can audit, agents can patch, MCP everywhere.
We mixed project brain with Mem0-lite facts: add_memory, search_memories, promote_memory_to_file, and optional extract — without turning into a full connector/RAG cloud.
{
"mcpServers": {
"readmedb": {
"url": "https://app.readmedb.com/api/mcp",
"headers": {
"Authorization": "Bearer rdb_your_key"
}
}
}
}Start free project brain →