MCP Redis Monitor

What antonio-mello-ai/mcp-redis-monitor MCP server does

The antonio-mello-ai/mcp-redis-monitor MCP server exposes read-only monitoring queries for a Redis instance. Its five tools cover queue contents, Celery queue state, client connections, server-level statistics, and key counts across Redis databases.

The server is intended for inspection rather than administration. The provided material describes no commands for setting, deleting, or modifying Redis data. That makes it suitable for asking an MCP-connected agent about current Redis workload and basic health indicators while keeping the server’s tool surface read-only.

How it works

The process runs locally and connects to Redis using configuration supplied through environment variables. Redis defaults to localhost on port 6379; a password can be supplied when the instance requires one. After startup, an MCP client can invoke the five exposed tools and receive monitoring results from the configured Redis instance.

get_queue_depths examines keys and their lengths by type in a selected database. Its documented default database is database 3. get_celery_queue_status reports Celery queue names, pending task counts, and the oldest task. get_connected_clients returns the number of connected clients.

get_server_info provides server memory information, uptime, and keyspace statistics. get_key_count_by_db gives an overview of key counts for each database. Together, these tools provide separate views of queue load, worker-related queue state, client activity, server status, and database usage.

Setup and configuration

The antonio-mello-ai/mcp-redis-monitor MCP server can run directly with uvx:

uvx mcp-redis-monitor

It can also be installed with pip and then started with:

pip install mcp-redis-monitor
mcp-redis-monitor

Configure the Redis connection with these variables:

• REDIS_MONITOR_HOST: Redis hostname, defaulting to localhost.
• REDIS_MONITOR_PORT: Redis port, defaulting to 6379.
• REDIS_MONITOR_PASSWORD: Optional Redis password.

An MCP client configuration can launch the server using the mcp-redis-monitor command. The README shows a server entry named redis-monitor with that command and no additional arguments.

Tools and capabilities

The available tools are:

• get_queue_depths: Lists keys and lengths by data type in a selected database; database 3 is the default.
• get_celery_queue_status: Reports Celery queue names, pending counts, and the oldest task.
• get_connected_clients: Reports the number of connected Redis clients.
• get_server_info: Returns memory, uptime, and keyspace statistics.
• get_key_count_by_db: Summarizes key counts by Redis database.

Limitations and notes

The antonio-mello-ai/mcp-redis-monitor MCP server is documented as read-only, so it should not be used when an agent needs to change Redis keys, queues, configuration, or server state. Results depend on access to the configured Redis instance and on the Redis data and Celery queues present there. The documented configuration covers one Redis host, port, and optional password; no other connection or authentication mechanisms are specified.

The project is built with FastMCP and is released under the MIT license. The README includes Ruff commands for linting and formatting and Pytest for running its test suite, but it does not document additional runtime services or named client integrations.