Redis Agent Memory is now available on the Snowflake Marketplace, adding persistent long-term memory to Snowflake Cortex AI agents. It runs inside a Snowflake account via Snowpark Container Services with no data egress. The solution provides four memory types — episodic, semantic, user preference, and procedural — accessible through SQL functions like store_memory(), search_memory(), and hydrate_prompt(). Installation takes under five minutes, requires no external infrastructure, and is Capacity Drawdown eligible using existing Snowflake credits.
Table of contents
The memory gap in Cortex agentsHow Redis fills the gapUp & running in five minutesAgents that actually rememberQuestions this post answers
How do I add persistent memory to Snowflake Cortex agents?
Redis Agent Memory, available on the Snowflake Marketplace, adds persistent memory to Cortex agents without external infrastructure. It runs inside your Snowflake account via Snowpark Container Services. It provides four memory types — episodic, semantic, user preference, and procedural — exposed as SQL functions: store_memory(), search_memory(), and hydrate_prompt(). Installation requires granting two privileges and auto-provisions four containers in under five minutes. Teams shipping Cortex agents to production track tools like this on daily.dev before committing to a memory architecture.
What does the hydrate_prompt() function do in Redis Agent Memory for Snowflake?
hydrate_prompt() is a single SQL function call that assembles relevant long-term memory — facts, history, preferences, and process steps — into a complete context payload for an LLM prompt. It replaces the manual work of searching, ranking, merging, and formatting context, so developers do not need to orchestrate those steps themselves before each model call. Developers choosing between memory architectures for Snowflake agents find the trade-offs covered on daily.dev.