Upstash Redis MCP. Manage cache state directly from your chat or terminal.

Listen up. The Upstash Redis MCP Server connects your agent straight to a serverless Redis cache. You don't gotta open some clunky database client or mess with REST API calls yourself; your AI client handles all that plumbing for you. It lets you treat the cache like it’s right in your terminal, letting you read and write key-value data structures without lifting a finger.

This is how you debug rate limits or manage session caches straight from your chat window.

Reading Data & Writing Values

If you need to grab a value, you use the get tool. You just point it at a specific key, and it pulls out the exact string stored there. Need to put something new in? The set tool handles that. You write a brand-new string value to a given key, and here's the trick: when you run set, you can also tack on an optional expiration time (TTL).

This means you tell it exactly how long that data needs to stay alive before Redis automatically wipes it clean. It’s perfect for temporary session tokens or cached API responses.

Inspecting the Cache State

You gotta know what's in there, right? To see if a key is doing its job, you run get_key_info. This tells you two things: first, what kind of data type it is (is it a string, or something else?), and second, how much Time-To-Live (TTL) it's got left. Need to know everything in the database? Use list_keys.

You feed it a glob pattern—like an asterisk matching anything—and it scans the whole thing, spitting out every key that matches your criteria. That lets you audit large sections of cached data fast.

Handling Counters and Maintenance

Sometimes you're not dealing with strings; you're counting things. The increment tool is for that. You use it to atomically increase or decrease a numeric counter tied to a specific key. This mechanism is killer for rate limiting logic or tracking how many times a user has hit an endpoint—it guarantees the count stays accurate, no matter what.

When you're done with data or need to clear out old garbage, delete does the dirty work. You specify which one or more keys are stale, and it wipes them from the cache entirely. It’s your clean-up crew. Before you trust anything, though, you run ping. This simple check confirms Redis is actually running and reachable, making sure your agent doesn't throw a fit when it tries to connect.

You use these tools together—get pulls data, set writes new data with expiration rules, list_keys finds everything matching a pattern, get_key_info checks the metadata and TTL, increment counts up or down, delete clears out what's stale, and ping confirms you’re live. That gives your agent full command over your cache layer.