[Feature Request] CrewAI Framework Integration for Memori Python SDK

Is your feature request related to a problem?

Summary

Hi Memori team

I'm Rachit Kakkad, a CS student and open source contributor (GitHub: Rachit-Kakkad1). I've been going through Memori's architecture and noticed CrewAI is missing from the supported frameworks list alongside Agno, LangChain, and Pydantic AI.

I'd like to build this integration and wanted to align with the team before starting.

---

Why CrewAI?

• 30M+ downloads, one of the most widely used multi-agent frameworks in the Python ecosystem
• CrewAI's default memory is machine-bound and session-scoped - it gets wiped on every redeployment or container restart
• The demand for external persistent memory in CrewAI is already proven (Mem0 built this integration, developers are actively searching for alternatives)
• Unlike other framework requests (#175 AgentScope), CrewAI has an established external memory provider pattern making the integration path well-understood and mergeable

---

Proposed Approach

Hook into CrewAI's agent lifecycle to automatically capture:

• Every agent tool call and its outcome
• Agent decisions and reasoning steps
• Conversation turns between agents
• Task completions and results

This maps directly to Memori's existing attribution + session model - each CrewAI Crew becomes a process, each user/run becomes an entity.

Scope:

• Python SDK integration (primary)
• Full test coverage following Memori's existing patterns
• Example added to the Memori Cookbook
• Docs
• TypeScript SDK (stretch goal)

---

Before I Start

1. Is this already in progress internally or assigned to someone?
2. Any architectural guidance on the best lifecycle hook points in the current SDK?
3. Any PR conventions I should follow beyond CONTRIBUTING.md?

Happy to jump on Discord too if that's easier to discuss.

Thanks for building Memori

— Rachit Kakkad GitHub: Rachit-Kakkad1

The Solution

I'd like to build a first-class Memori x CrewAI integration that hooks into CrewAI's agent lifecycle - automatically capturing every tool call, agent decision, and conversation turn into Memori's structured memory layer.

Each CrewAI Crew maps to a Memori process, each user/run maps to an entity. This fits directly into Memori's existing attribution + session model with zero changes to the core SDK.

Scope:

• Python SDK integration (primary)
• Full test coverage following Memori's existing patterns
• Example added to the Memori Cookbook
• Docs
• TypeScript SDK (stretch goal)

Alternatives Considered

Mem0 already built a CrewAI integration and developers are actively using it - meaning the demand is proven and the external memory provider pattern in CrewAI is well understood. Memori needs this to stay competitive in the multi-agent framework space.

Affected Components

Other

Participation

• I am willing to submit a pull request for this feature.