claude-mem hits 46.1K stars as persistent memory plugin for Claude Code

One of the most consistent complaints I hear from developers using AI coding tools is that they forget everything between sessions. Every new conversation starts cold. You re-explain the project structure, the recent decisions, the ongoing bugs -- and then do it all again tomorrow.

claude-mem is the most practical solution to that problem I've come across. The open-source plugin automatically captures what Claude does during coding sessions, compresses observations with AI, and injects relevant context into future sessions. It just crossed 46.1K GitHub stars, which tells me this pain point is much more widespread than vendors would like to acknowledge.

What Happened

Developer Alex Newman (@thedotmack) released v12.0.0 of claude-mem, the project's latest release across 223 total releases. It has grown to 92 contributors and 3.5K forks since its initial launch, which is a lot of iteration for a plugin.

The tool hooks into Claude Code's session lifecycle, capturing tool usage, generating semantic summaries via Claude's agent-sdk, and storing everything in a local SQLite database with Chroma vector search. Installation is a single command: npx claude-mem install.

What I'd flag here is the breadth of support. It also works with Gemini CLI and the OpenClaw gateway, which means it's not locked to Anthropic's ecosystem. For teams hedging across providers, that matters.

Key Features

• Automatic session capture. Five lifecycle hooks (SessionStart, UserPromptSubmit, PostToolUse, Stop, SessionEnd) record what Claude does without any manual intervention. You don't have to remember to log anything.
• AI-compressed memory. Observations are compressed and summarized using Claude's agent-sdk, then stored in SQLite with FTS5 full-text search. The compression is what keeps token costs manageable.
• Hybrid vector search. Chroma vector database enables semantic and keyword search across your project history. Local embeddings use all-MiniLM-L6-v2 via ONNX, so no external API calls are needed.
• Token-efficient retrieval. A 3-layer MCP tool workflow (search, timeline, get_observations) claims roughly 10x token savings by filtering before fetching full details. That's the kind of efficiency that matters on long-running projects.
• Privacy controls. <private> tags let you exclude sensitive content from storage. All data stays local by default.
• Multi-IDE support. Works with Claude Code, Cursor, Gemini CLI, Windsurf, and OpenClaw gateways, cross-platform on Linux, macOS, and Windows.

Why It Matters

The problem is simple and annoying: AI coding assistants forget everything between sessions. claude-mem attacks this directly rather than working around it.

For teams running long-lived projects, persistent memory means Claude can reference yesterday's debugging session or last week's architecture decision without being told. The multi-machine sync feature (claude-mem-sync) even lets developers share observations and session summaries between machines over SSH, which is the kind of detail that tells me this was built by someone who actually uses it on real projects.

The broader pattern I'm seeing here is that as AI coding tools mature, the bottleneck shifts from model capability to context management. Developers don't just need smarter models. They need models that remember. claude-mem is one of the more concrete expressions of that shift.

Example Use Case

The problem is simple and annoying: AI coding assistants forget everything between sessions. claude-mem attacks this directly rather than working around it.

For teams running long-lived projects, persistent memory means Claude can reference yesterday's debugging session or last week's architecture decision without being told. The multi-machine sync feature (claude-mem-sync) even lets developers share observations and session summaries between machines over SSH, which is the kind of detail that tells me this was built by someone who actually uses it on real projects.

The broader pattern I'm seeing here is that as AI coding tools mature, the bottleneck shifts from model capability to context management. Developers don't just need smarter models. They need models that remember. claude-mem is one of the more concrete expressions of that shift.

Competitive Context

Claude Code ships with CLAUDE.md files for project context, but they're static. Developers maintain them manually. claude-mem automates this by generating and updating folder-level CLAUDE.md files from its observation database, preserving any manually written content in the process.

Other memory solutions exist in the broader AI tooling space -- mem0, supermemory, openmemory -- but claude-mem is purpose-built for Claude Code's plugin architecture. It hooks directly into session lifecycle events rather than wrapping API calls. That's a meaningful architectural difference, and it's what makes the context injection feel native rather than bolted on.

For Gemini CLI users, claude-mem provides the same persistent memory layer, making it one of the few tools that spans both Anthropic and Google's coding environments. That cross-ecosystem support is underrated for teams that don't want to commit entirely to one vendor.

My Take

claude-mem solves a specific, painful problem cleanly. If you use Claude Code on projects that span more than a single conversation, this is worth installing. The 46.1K stars reflect real utility, not hype -- context management is quietly becoming the most important layer in AI-assisted development, and claude-mem is one of the better tools addressing it.