RAMen
An in-memory data store that speaks the Redis wire protocol as a drop-in cache, while adding native vector search, semantic caching of AI answers by meaning, and a built-in MCP server so agents can use it as a tool directly.
RAMen behaves like Redis for common caching and key-value use cases, since it speaks the same RESP2 protocol, so an existing Redis client and code can point at RAMen with no changes for basic get, set and cache operations. Beyond that compatibility, it adds two capabilities Redis needs an extra module for: built-in vector storage and search, and a semantic cache that can recognize when two differently worded prompts mean the same thing and return the previously computed answer instead of calling a model again.
A built-in MCP server lets AI agents such as Claude read, write, search and remember data in RAMen directly as a tool, with no separate glue code. It ships as a single small binary with no external dependencies, installable with Go, Docker or from source, and the project states plainly what it does not yet do: no clustering, replication or failover, no append-only-file durability, since it saves snapshots to disk instead, and it is not tuned to beat Redis or Valkey on raw throughput. It is written in Go and BSD-3-Clause licensed.