Smara Memory API

Smara Memory API provides persistent, cross-tool memory for AI tools, addressing the challenge of AI assistants forgetting user preferences and context between sessions. It enables AI tools such as Claude Code, Cursor, Windsurf, Codex, and VS Code to share and recall contextual information, ensuring that relevant facts and decisions persist and are available regardless of the tool or session. This shared memory decays naturally over time, prioritizing frequently used facts and allowing less relevant information to fade, modeled after Ebbinghaus decay principles.

The platform operates via a simple REST API with three core endpoints: storing new facts, searching memories, and retrieving user context. It requires no SDKs or schema definitions, making integration straightforward for developers working in any language. Smara automatically captures context as users work, eliminating the need for manual tagging or commands. The memory pool is accessible across supported AI tools, so context follows users as they switch between applications. Facts are ranked by both relevance and freshness, and the tool provides visibility controls, allowing memories to be marked as private or shared with a team.

Smara Memory API is available as an open-source project under the MIT License. Users can self-host the service with Docker or access a free hosted API, with no credit card required to get started. The tool is designed for teams and individuals seeking to enhance the continuity and usefulness of their AI-driven workflows without repetitive re-explanation of preferences or project details. Installation involves adding Smara to an MCP configuration, and it works with any MCP client.

The platform demonstrates high accuracy on the LoCoMo benchmark and provides a playground, documentation, and GitHub repository for further exploration. By enabling persistent, decaying memory across multiple AI tools, Smara addresses a key limitation in current AI assistant workflows.