Launching today
Cache-Pot
A superfast memory layer built for AI agents
7 followers
A superfast memory layer built for AI agents
7 followers
Cache-Pot is a Redis-compatible in-memory data store, rebuilt for how AI apps and agents use a cache. Speak plain RESP2, your existing Redis client works untouched. On top: native vector search, a semantic cache (skip repeat LLM calls when a question means the same thing), a built-in MCP server so agents like Claude use it as a first-class tool, and a live web dashboard. All in one self-contained binary, no modules, no extra services.
CodeNearby
One thing that would make this even more useful for me: built-in TTL policies tied to semantic similarity scores, so near-duplicate entries auto-expire before drifting too far from the original. Right now the semantic cache seems to grow forever unless I set a hard TTL manually, and that can lead to stale answers sneaking through.
Dropped my existing redis client at it and everything just worked, which never happens. The semantic cache alone has already cut a chunk of repeat calls from my agent's loop.