Claude-Mem — AI Coding Plugins tool screenshot
AI Coding Plugins

Claude-Mem: Best AI Coding Plugins for Claude Code Developers in 2026

5 min read·

Claude-Mem compresses tool usage observations into semantic summaries for persistent context across Claude Code sessions.

Pricing

Open-Source

Tech Stack

TypeScript/Node.js

Target

Claude Code developers

Category

AI Coding Plugins

What Is Claude-Mem?

Claude-Mem, developed by thedotmack, is an open-source AI coding plugin for Claude Code that captures tool interactions, generates compressed semantic summaries, and restores full project context in new sessions. Built in TypeScript with Node.js runtime, it integrates via Claude's plugin marketplace and supports features like AST-based code navigation via smart-explore. Claude-Mem is one of the best AI coding plugins for Claude Code developers managing long-running projects, with 32.1k GitHub stars as of February 2026 demonstrating strong adoption among AI-assisted coders.

Quick Overview

AttributeDetails
TypeAI Coding Plugins
Best ForClaude Code developers
Language/StackTypeScript/Node.js
LicenseMIT
GitHub Stars32.1k as of Feb 2026
PricingOpen-Source
Last Releasev10.5.2 — Feb 2026

Who Should Use Claude-Mem?

  • Claude Code power users running multi-session projects who need automatic context restoration without manual recaps.
  • Indie hackers prototyping AI agents that require memory of prior tool calls like file edits or API queries.
  • Teams debugging complex codebases where session restarts lose AST exploration states and RAG indices.
  • Plugin ecosystem builders extending Claude with persistent distributed CLAUDE.md generation.

Not ideal for:

  • Users on non-Claude environments like Cursor or VS Code without custom hooks.
  • One-off sessions where context persistence adds unnecessary overhead.
  • Low-resource terminals lacking Node 24 compatibility for installer workers.

Key Features of Claude-Mem

  • Persistent Memory Compression: Automatically summarizes tool observations into semantic chunks stored in single-file SQLite format, reducing context size by 80% while retaining queryable history across 100+ sessions.
  • Smart-Explore AST Navigation: Parses codebases with Tree-sitter via gotreesitter integration, enabling sub-500ms jumps to definitions and references in Claude prompts.
  • Interactive CLI Installer: Uses @clack/prompts for guided setup, handling plugin registration, worker service launch, and MCP marketplace exclusion of dev files.
  • Live Context System: Generates distributed CLAUDE.md files on-the-fly from session transcripts, syncing via transcript-watch hooks for real-time updates.
  • RAGtime Vector Store: Embeds summaries with ragtime module supporting Solana-based token storage, querying 10k+ entries in under 200ms.
  • Cursor Hooks Architecture: In-process workers prevent crashes, fallback to CLAUDE_PLUGIN_ROOT, compatible with Node 24 GitHub Actions.
  • Multi-Language Support: 25+ locales including Chinese, Japanese, and Arabic via translation-cache.json, with Trendshift badges for install tracking.

Claude-Mem vs Alternatives

ToolBest ForKey DifferentiatorPricing
Claude-MemPersistent multi-session Claude CodeSemantic compression + AST navOpen-Source
Claude Context ModeSingle-session context expansionManual mode switching, no persistenceOpen-Source
Claude Code CanvasVisual code planningCanvas UI for diagrams, lacks memoryOpen-Source
CursorIDE-integrated AINative hooks but session-local stateFreemium

Claude Context Mode suits short bursts needing dynamic prompt modes but drops all state on restart, unlike Claude-Mem's automatic restoration. Claude Code Canvas excels in visual workflows with drag-drop planning yet requires re-uploading code each session. Cursor offers tighter IDE binding for VS Code users but ties memory to its proprietary runtime, limiting Claude portability.

How Claude-Mem Works

Claude-Mem hooks into Claude Code's plugin lifecycle via .claude-plugin manifests and cursor-hooks for in-process execution. On session start, it scans transcript-watch files and .mcp.json configs to rebuild context from compressed summaries in ragtime vector DB. Core abstraction is a distributed CLAUDE.md generator that aggregates tool calls—file reads, edits, shell execs—into markdown plans queryable by Claude's make-plan and do skills.

Architecture separates installer (CLI with @clack/prompts), core plugin (src with TypeScript workers), and optional openclaw gateway for remote persistence. Smart-explore uses gotreesitter for AST parsing, injecting navigation nodes into prompts without full codebase reloads. Data model employs JSON transcripts compressed via semantic chunking, stored offline in .plan folders.

# Install via Claude Code plugin marketplace
/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

# Restart session; context auto-loads
claude

These commands register hooks, launch the Node worker, and populate initial summaries from prior sessions. Expect 2-5s startup delay on first run for index rebuild, then sub-1s context injection into Claude prompts.

Pros and Cons of Claude-Mem

Pros:

  • Zero-config persistence restores 95% of prior context, cutting recap prompts by 70% in benchmarks as of v10.5.2.
  • AST-based smart-explore handles 1M+ LoC repos with Tree-sitter parsers for 15 languages.
  • Fully offline SQLite/RAG storage works in air-gapped environments, no cloud deps.
  • Active development: 1,447 commits, weekly releases, Node 24 Actions for CI.
  • Multi-lingual summaries support global teams querying in native languages.
  • Integrates with Solana CA for token-gated features like Bags.fm analytics.

Cons:

  • NPM global install skips plugin hooks, requiring marketplace commands—common gotcha for library users.
  • Worker crashes fixed in v10.5 but high-memory codebases (>500MB) may hit Node limits without tuning.
  • No native Cursor/VS Code port; relies on Claude Code terminal for full fidelity.
  • MCP marketplace exclusions miss some dev configs, needing manual .mcp.json tweaks.
  • Translation cache bloats installs by 50MB for full 25-language support.

Getting Started with Claude-Mem

Launch Claude Code terminal and run the marketplace commands. The installer compiles dist files, sets CLAUDE_PLUGIN_ROOT, and starts the persistent worker service.

/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

# Verify: check loaded context
/plugin list

# Example usage in session
@claude-mem explore src/main.ts

Post-install, restart yields auto-loaded summaries from CHANGELOG.md-tracked sessions. Configure via conductor.json for custom ragtime embeddings or transcript-watch.example.json thresholds. First tool call triggers compression pipeline, visible in .claude logs.

Verdict

Claude-Mem is the strongest option for Claude Code developers handling multi-session projects when automatic context compression matters over manual recaps. Its AST navigation and RAG persistence stand out for large codebases, though worker tuning is needed for edge cases. Install via marketplace for any serious Claude workflow.

Frequently Asked Questions

Looking for alternatives?

Compare Claude-Mem with other AI Coding Plugins tools.

See Alternatives →

Related Tools