Redis ships an official MCP server that gives AI agents a natural-language interface to your Redis instance — every core data structure, plus search and vector queries. It runs locally over stdio and connects with a standard Redis URL.
Bring your own
Claude Sonnet 4.5
Chat with 60+ AI models on the same workflow — switch to a different model mid-conversation and re-run the same prompt, or use Compare mode to put several side-by-side and balance quality vs. cost.
A Redis connection URL, typically redis://user:password@host:port. Use a restricted ACL user rather than the default one.
How models use it and what it is built for.
The Redis MCP server maps Redis operations onto MCP tools so an agent can work with your data without you writing commands. It covers the full type system — strings, hashes, lists, sets, sorted sets and streams — along with key management, TTL inspection and pattern-based scanning. Where it gets more interesting is search: on instances with the query engine enabled, an agent can run full-text and vector similarity queries, which makes it a practical backend for retrieval workflows. A separate server, mcp-redis-cloud, covers the Redis Cloud control plane if you want an agent managing databases and subscriptions rather than data.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
How many keys match session:* and what is the average TTL left on them?
Show me the top 10 entries in the leaderboard sorted set with their scores.
Read the last 20 messages from the events stream and summarise what happened.
Find the five documents most similar to this embedding in the products index.
This is not a single-model product: you get the same MCP connection with 60+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Redis
Claude Sonnet 4.5
Sonnet 4.5 picks the right data-structure command from a plain question and knows to SCAN rather than KEYS on a large instance.
Listing tools proves the server is reachable, not that a model can work with it. Evals go further: they read every tool on the server, write a test suite from its real schemas, and run it — code decides pass/fail on the responses (schema conformance, error codes, pagination, result caps) while a scoring model grades plain-English tasks driven through the tools.
Get a pass/fail report per tool with the evidence behind each verdict — and replay the same suite after every schema change. Destructive tools are excluded from the run.
Run evalsOpen MCP Agent Studio with the connection pre-filled. Add your token, pick any of 60+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
It is the official MCP server from Redis, a natural-language interface for agentic applications to manage and search data in Redis. It covers hashes, lists, sets, sorted sets and streams, plus search and vector queries on instances with the query engine enabled.
mcp-redis operates on the data inside a Redis instance — reading and writing keys. mcp-redis-cloud operates on the Redis Cloud control plane, so an agent can manage subscriptions and databases. Pick the first for data work and the second for provisioning.
Use a dedicated ACL user rather than the default. Redis ACLs let you restrict both commands and key patterns, so you can grant read-only access to one namespace. That is a much better control than trusting the agent to avoid destructive commands.
Yes, on instances that have the Redis query engine enabled. The agent can run vector similarity and full-text queries against indexes you have already created, which makes it usable as a retrieval backend inside a larger agent workflow.
A capable model will reach for SCAN on a large keyspace, but this is exactly why you should point it at a replica or a scoped ACL user first. Test the behaviour in MCP Agent Studio, where every command the model issues is visible before you trust it against production.