Memmy

Purpose

Memmy is a local-first AI agent positioned to solve the problem of context loss when users switch between different AI tools. Through a unified personal memory layer, it allows desktop applications, command-line interfaces (CLI), and external AI agents to share the same long-term context, thereby achieving "letting every AI remember the same you."

Main Capabilities

• Unified Personal Memory: Share preferences, decisions, and progress across the multiple AI tools used by the user.
• Remember and Understand: It can remember what the user has said (for example, "Use Tauri for this project, not Electron") and understand the user's intent (for example, "I mentioned my style before - concise and restrained"), so there is no need to repeat explanations in subsequent interactions.
• Searchable Memory: After obtaining the user's permission, read supported AI conversation history on the device and organize useful details such as preferences, decisions, and project context into searchable memories.
• Continuity of Context Across Tools: When switching among tools such as Cursor, Claude Code, and Codex, there is no need to rebuild context; Memmy carries relevant decisions, project history, and prior work.
• Agent Tasks: Users can directly assign tasks to Memmy, and it uses relevant memories to organize materials, integrate plans, or turn past decisions into documents.

How to Use

1. Authorized Scan: After the user grants permission, Memmy scans the conversation history of local AI tools (including Cursor, Claude, and Codex).
2. Organize Memories: Memmy turns scattered conversations into searchable memories, covering preferences, decisions, project context, and past lessons learned.
3. Invoke on Demand: When the user uses connected AI tools, Memmy introduces only memories relevant to the current task, rather than the entire history.

Applicable Scenarios

• Developers who frequently switch among multiple AI coding assistants (such as Cursor, Claude Code, and Codex).
• Users who want AI to remember personal preferences and project decisions and avoid repeated explanations.
• Scenarios that require organizing past decisions and project history into documents or plans.

Limitations

• Memories are stored locally by default, and users can manage access permissions and data, but the specific AI tools and operating system versions supported are only explicitly mentioned in the materials as macOS 12+, Windows 10+, and Cursor, Claude, Codex; other platforms or tools are not yet provided.
• Reading history requires explicit user authorization, and users can stop further collection at any time.
• The product is open source and free, but no detailed information is provided on charging models, pricing, or specific feature limitations.