Happy slice of mcp pizzaMCP.pizza

agent-memory-server

MCP.Pizza Chef: redis

Built by Redis, it holds what happened in a conversation and quietly promotes the important details into longer-term memory, so an assistant can recall your preferences and past decisions days later. The free code now sits in a folder marked V0, which Redis describes as a research foundation rather than the supported path; the company points teams to its hosted Agent Memory service in Redis Iris instead. Running the open version yourself means standing up Redis and doing real installation work.

Deprecated · Redis points teams to its hosted Agent Memory service in Redis Iris
Coding
Data

Use This MCP server To

Have my assistant remember my preferences between sessions Recall a decision we made in a chat last week Keep a long conversation from losing its earlier context Search past conversations by meaning rather than exact words Set how long a conversation's memory is kept

README

Redis Agent Memory

A memory layer that gives agents intelligent short-term memory and persistent context across conversations.

Redis Agent Memory in Redis Iris

Redis Agent Memory in Redis Iris is Redis’s official managed path for teams that want agent memory as a service, not another subsystem to build and operate themselves. Redis Iris is the real-time context engine for agents, designed to deliver fresh, relevant context at runtime, and Redis Agent Memory is the part of Iris that makes context compound across turns, sessions, channels, and agents.

Redis Agent Memory in Iris gives you the Redis-managed experience: a persistent, structured memory layer for AI agents exposed through a REST API and client libraries, with dedicated endpoints, secure API key management, configurable memory schemas, and automatic TTL-based lifecycle management. The point is not just storage. It is to remove the custom memory infrastructure teams otherwise end up building around session handling, extraction, retrieval, and lifecycle management.

Redis Agent Memory uses a two-tier model. Session memory keeps the active conversation state, session history, and session-specific metadata close at hand, with configurable TTL control for retention. Long-term memory stores extracted facts and learned patterns from past interactions as text plus vector embeddings for semantic retrieval. As new events are written to working memory, Redis Agent Memory automatically extracts important information and promotes it to long-term memory in the background, so memory accumulates without slowing down the live agent loop.

That matters because Redis Iris is not just a memory feature in isolation. It is a broader context engine built to address the production problems agents actually hit: fragmented data, stale operational state, slow retrieval, and interactions that do not improve over time. Within that story, Redis Agent Memory is the compounding memory layer; Redis Context Retriever makes business data navigable; Redis Data Integration keeps operational state fresh; and Redis LangCache helps repeated work stay inside the latency budget.

If you are evaluating the supported Redis path, these are the best places to start:

A practical getting-started flow on Redis Cloud looks like this:

For implementation details and usage examples, see:

V0 — the open-source research foundation

V0/ contains the original Redis Agent Memory Server: an open-source reference implementation for agent memory with REST and MCP interfaces, working and long-term memory, configurable extraction strategies, and Redis-backed semantic search. It serves as the research foundation and architectural starting point for Redis Agent Memory, but it is not the current supported production path.

License

This project is licensed under the Apache License 2.0 (Redis, Inc.). See LICENSE at the repository root.

agent-memory-server FAQ

Is this the current version?
Not any more. The free code has been moved into a folder named V0 and Redis now calls it a research foundation, pointing teams to its paid Agent Memory service in Redis Iris for production use.
Can I use this to make my assistant remember me between chats?
Yes. That is the whole point of it — details from one conversation are saved and can be recalled in a later one.
How hard is setup?
Hard. You have to run Redis and the memory service yourself, so this is developer territory rather than a setting you paste in.
Do I need to pay for anything?
The free version needs a key for whichever model does the summarising, and the hosted Redis version is a paid product.
Which apps does it work in?
It is aimed at people building their own assistants rather than users of one particular chat app, though the free version does offer a standard connection.
Where does my conversation data live?
If you run the free version yourself, it stays in the Redis instance you control.