Analysis updated 2026-05-18
Write AI application memory logic once and switch between memory backends like Supermemory or Letta by swapping an adapter.
Validate memory-related requests at runtime using the included Zod schemas.
Build a new adapter for another memory backend so it works with existing app code that targets the shared interface.
Run conformance tests that check a memory backend behaves the same as Supermemory's documented API.
| wazootech/memsdk | 0xradioac7iv/tempfs | 7vignesh/pgpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Installed directly from GitHub rather than npm, and needs a specific memory backend adapter such as the existing Letta one to actually run against a service.
memsdk is a TypeScript library that gives AI applications a single, consistent way to talk to different memory backends. Memory backends are services that store and retrieve information an AI assistant needs to remember, such as past conversations or documents. The problem this project addresses is that each memory backend, such as Supermemory, Letta, or Mem0, ships its own SDK with a different shape, so switching between them normally means rewriting an app's memory handling code from scratch, even though the basic actions like adding, searching, forgetting, and listing items are conceptually the same everywhere. Rather than inventing a brand new universal interface, which the author notes has already been tried and abandoned elsewhere, memsdk instead takes Supermemory's existing API surface, which the author considers the most complete one available, and turns it into a backend-agnostic contract that any memory service can implement. This contract is defined as TypeScript types plus Zod schemas for validating requests at runtime, with no other runtime dependencies. An application is written once against this shared interface, and different memory backends can then be swapped in by writing a small adapter for each one, without changing the application code itself. A working adapter for Letta already exists in a companion repository, along with a separate end-to-end test suite that runs the same set of scenarios against both Supermemory and Letta side by side to confirm they behave the same way. The current version covers documents, search, memories, and profile actions, while things like settings, connections, and raw HTTP helpers are intentionally left out for now. The package is installed directly from GitHub rather than from the npm registry, and it works with Node.js, Bun, pnpm, yarn, and bundler-based browser projects such as Vite, Deno and direct browser or CDN usage are not yet supported.
memsdk is a TypeScript library that lets developers write AI memory-handling code once against a shared interface, then swap between different memory backends like Supermemory and Letta without rewriting app code.
Mainly TypeScript. The stack also includes TypeScript, Zod, Node.js.
No license information is provided in the README.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.