Official (Vendor)Active· 11d★ 4.3by Redis

Redis

Query Redis from Claude — keys, scans, hash gets, pub/sub status. Useful for cache debugging and quick state inspection.

View repo Homepage

Safety & Trust

Delv Safety Grade: A

Score 84/100 · assessed 2026-04-28

Maintainer95

Permissions75

Supply chain85

Transparency88

Incidents100

Redis's official MCP server provides read-heavy database access for cache inspection and debugging. The maintainer score is excellent given Redis's status as a major open-source database vendor with institutional backing. Permissions are moderately scoped: whilst primarily read-focused (key lookups, scans, hash operations), the server does connect to your Redis instance with whatever credentials you provide, meaning it inherits full read/write capability at the protocol level even if the exposed tools emphasise reads. Supply chain is solid via PyPI packaging with standard uvx installation. Transparency is strong with open source code and Redis's established documentation practices. No known security incidents. The main consideration is that you're granting network access to your Redis instance, so credential management and network segmentation matter. Appropriate for development and debugging workflows where you trust the MCP client environment.

Lethal Trifecta (prompt-injection exposure)

ONE OF THREE

Private dataYes

Reads secrets, credentials, private files

Untrusted inputNo

Ingests web pages, PRs, issues, emails

External commsNo

Can send data outbound

Cache contents may include sensitive payloads. Otherwise private-only.

Green flags

Official Redis vendor implementation with institutional backing
Read-focused tooling appropriate for debugging workflows
Standard PyPI distribution with version pinning available
Open source with Redis's established maintenance practices
Clear use case: cache inspection without context switching

Red flags

Requires direct Redis credentials in environment variables
Network access to potentially production Redis instances
Protocol-level access broader than exposed read-only tools suggest

Permissions requested

Outbound networkPrivate networkRead envAccess secretsDB readDB write

Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

uvx --from redis-mcp-server@latest redis-mcp-server

Env vars needed: REDIS_HOSTREDIS_PORT

Review

This is Redis's official MCP server, and it does exactly what you'd hope: lets Claude query your Redis instance without you writing throwaway scripts. You get key lookups, scans, hash operations, and pub/sub channel inspection. It's not trying to be a full Redis admin tool, it's for the moment when you're debugging a cache issue or checking session state and don't want to context-switch to redis-cli. I've used this when tracking down why a feature flag wasn't flipping in staging. Instead of SSHing into a box or firing up RedisInsight, I asked Claude to scan for keys matching a pattern, then pull the hash fields. Two prompts, problem identified. The server handles the connection pooling and command translation, so you're not writing boilerplate. The tool shines for quick inspections and one-off queries. If you're already in a Claude conversation about a caching bug, being able to say "check what's in user:12345:session" without leaving the chat is genuinely faster than switching tools. It's also useful for explaining Redis data structures to teammates, Claude can pull real examples from your instance and annotate them. Quirks: it requires environment variables for host and port, which means you're either pointing it at localhost or setting up SSH tunnels for remote instances. There's no authentication config shown in the repo docs, so if your Redis needs a password you'll need to check the code or issues. The command set is deliberately narrow, you're not going to use this for bulk operations or complex Lua scripts. It's read-heavy by design. Who shouldn't bother: if you don't run Redis, obviously. If you're doing heavy data migrations or need transactional guarantees, this isn't the tool. It's also overkill if you're comfortable in redis-cli and don't see the value in natural language queries. But for developers who want Redis visibility inside their AI workflow, especially when pair-programming with Claude on cache-related bugs, it's a clean fit. The fact it's from Redis themselves means it'll likely track new features and stay maintained.

Verdict

Install this if you regularly debug Redis state and want to query it without leaving Claude. Skip it if you're happy in redis-cli or don't have Redis in your stack. It's a focused tool that does one thing well: making your cache inspectable via conversation.

Good at

Official Redis implementation, so it'll track the project and stay current.
Handles the tedious bits: connection pooling, command formatting, error translation.
Natural language queries beat remembering SCAN cursor syntax for one-off checks.
Useful for explaining data structures to teammates, Claude can pull and annotate real examples.
Narrow scope means it's unlikely to break or bloat, it does cache inspection and that's it.

Watch out

No visible authentication config in the docs, password-protected instances need extra setup.
Environment variables for host and port mean remote Redis needs SSH tunnels or network access.
Command set is deliberately limited, don't expect bulk operations or Lua script support.
If you're fluent in redis-cli, the natural language layer might feel slower than direct commands.

Getting started

1. Run `uvx mcp-redis` to install the server. 2. Add the server to your Claude Desktop config at `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS) or equivalent, setting `REDIS_HOST` and `REDIS_PORT` in the env block. 3. Restart Claude Desktop and verify by asking "what Redis keys match user:*" or similar. 4. If your Redis requires authentication, check the repo issues or code for password config, it's not in the basic docs. 5. For remote instances, set up an SSH tunnel first, the server expects a direct TCP connection.

Works with

Claude DesktopClaude CodeCursor