Smara

Gives an assistant persistent cross-platform memory by storing, searching, and deleting user facts through the Smara Memory API.

What you can do

This server connects an MCP client to the Smara Memory API so the assistant can keep long-term memory about a user. Tools include store_memory for saving a fact with an importance score, search_memories for semantic search using the vendor's Temporal Memory Scoring ranking, get_user_context for a pre-formatted context string to place in a system prompt, delete_memory for removing a specific memory, and get_usage for checking plan limits and memory count. The vendor says memories fade over time based on importance and access, and that contradictions are detected and resolved.

When to use it

Worth adding when you want an assistant to remember facts about a user across sessions and across different MCP-compatible clients, rather than losing context each time. It suits setups where recall quality matters and you are comfortable storing those facts with a third-party memory service.

Requirements

Runs locally over stdio, typically launched with npx, so Node.js and network access are needed. A required SMARA_API_KEY must be supplied as an environment variable; the optional SMARA_API_URL defaults to the vendor's API endpoint. A free API key is obtained from the vendor's site.