Mengram — Human-like Memory Backend for CrewAI (PR #4595)

Showcase
, ,
1

Hey everyone! I built a memory backend for CrewAI powered by mengram.io and submitted a PR (#4595).

What it does:

  • Drop-in replacement for CrewAI’s Memory — just pass Crew(memory=mengram_memory)

  • Three memory types: semantic (entities + knowledge graph), episodic (events with outcomes), procedural (learned workflows)

  • Server-side extraction — no local LLM or embedder needed

  • Free tier available

Quick example:

python

from crewai.memory.storage.mengram_storage import MengramMemory, MengramConfig

memory = MengramMemory(MengramConfig(api_key="om-..."))
crew = Crew(agents=[...], tasks=[...], memory=memory)

PR: github.com/crewAIInc/crewAI/pull/4595

Would love any feedback!

2

Interesting.. Why would our community use it? Can you please give us context.

3

@alibaizhanov Can you explain with Simple use case? I thought CrewAI memory management is quite robust and comprehensive. Just curious to know what the feature you are adding here.