Agent Memory Tools

Pulls version-specific documentation and code examples directly into LLM prompts, solving documentation staleness. MCP server and library integration.

Packs entire repositories into AI-friendly single-file formats. Supports custom output formats, token counting, and .gitignore-aware filtering.

Converts any Git repository into a prompt-friendly text digest optimized for LLMs. Web UI and CLI available.

CLI tool to convert codebases to LLM prompts with token counting, Handlebars templates, and tree-sitter filtering.

Python version of code2prompt with similar codebase-to-prompt capabilities.

Concatenate files into a single prompt suitable for LLMs. Simple, Unix-philosophy approach by Simon Willison.

Project-level instructions and context for Claude coding agent

Emerging universal standard (Feb 2026+) for agent instructions

Legacy project rules for Cursor IDE

New MDC-format rules for Cursor IDE

Project context for Google Gemini CLI agent

Code conventions and standards file

Persistent memory extension for Claude Code. Stores and retrieves context across coding sessions.

Temporally-aware knowledge graph engine for real-time context synthesis. Bi-temporal model tracking event occurrence and ingestion time.

Memory and context engine for AI applications. Benchmarked #1 on three memory benchmarks.

Query-based memory database for coding agents. Efficient storage and retrieval of coding context.

Encrypted persistent memory for coding agents with privacy-preserving retrieval.

Efficient lifelong memory via semantic lossless compression. Three-stage pipeline: compression, synthesis, retrieval.

ByteDance's open-source context database for AI agents. Unifies management of context (memory, resources, skills) through viking:// filesystem paradigm with L0/L1/L2 tiered loading and self-evolution.

Context bridge layer managing skills, scripts, and context for AI coding agents. Integrates with OpenViking for persistent team-wide memory.

Agent Skills as a Memory Layer — treats skills as persistent, queryable context.

Memory Operating System treating context as OS-level resource. Unifies skills, memory, and resource management.

Memory and context engine with vector-based retrieval. Triple benchmark leader.

Serverless, single-file memory layer. Packages data, embeddings, indexes, and metadata into portable .mv2 files with hybrid BM25 + vector search.

Fast, flexible agent memory using Redis vector similarity search with semantic and keyword retrieval.

Memory and information retrieval server for LLMs built in Rust. Incremental summarization with MAX_WINDOW_SIZE management.

Open-source memory layer for autonomous agents with automatic memory updates during interaction.

ChromaDB + Postgres memory with DBSCAN clustering for semantic organization.

Entity-centric knowledge graphs with thematic clusters and LLM-precomputed community summaries. Multi-hop graph-based reasoning.

Simple and fast RAG using dual-level knowledge graphs. 80ms retrieval vs 120ms standard RAG.

Temporally-aware knowledge graph engine with bi-temporal model. 94.8% DMR benchmark accuracy.

Knowledge engine enabling knowledge graph construction in 6 lines of code. Multi-backend.

Graph-native memory for conversation-to-knowledge-graph pipelines.

Neurobiologically inspired long-term memory using hierarchical graphs with structure-aware retrieval.

Universal memory layer for AI agents. Framework-agnostic design with multi-tier memory (short, medium, long-term).

Platform for stateful agents with three-tier memory (Core, Recall, Archival). OS-inspired architecture where agents run inside memory.

Personal memory and context framework for AI agent personalization.

High-performance drop-in Mem0 replacement with semantic deduplication, multimodal video reasoning.

Long-context optimization via end-to-end RL. Extrapolates from 8K to 3.5M tokens with <5% performance loss.

Memory operating system for personalized agents. Hierarchical storage with Storage/Updating/Retrieval/Generation layers.

Open-source LLM application platform combining Backend-as-a-Service with LLMOps. Visual RAG/workflow orchestration with integrated memory.

Modular RAG/agent framework with pluggable memory backends. LangGraph adds graph-based orchestration, LangMem provides structured memory.

Multi-agent framework with role-based architecture. Agents assigned to software engineering roles coordinated via message passing.

Multi-agent conversation framework with persistent session memory and human-in-the-loop workflows.

Workflow automation platform with AI agent capabilities. 600+ integrations.

All-in-one AI platform with RAG, memory, and workspace management.

Official reference memory implementation using knowledge graph. Stores entities, relations, and observations.

Temporal knowledge graph MCP server with Neo4j integration and bi-temporal modeling.

Universal memory layer with MCP integration. Multi-tier memory persistence.

GraphRAG memory server with customizable ingestion and search. Part of Cognee platform.

Agentic Memory framework exposed as MCP server.

Enhanced version of official MCP memory with additional features.

AI-driven development platform with agent autonomy for software engineering tasks. Comprehensive state management.

Pioneering agent architecture with episodic memory, reflection, and planning. Agents simulate believable human behavior.

Software engineering agent with shell/editor interaction for autonomous GitHub issue fixing.

OS-inspired cognitive architecture with tiered memory management.

LLM controller coordinating expert models from HuggingFace Hub.

Minecraft agent with external skill library and lifelong learning.

Enterprise agent memory layer with confidence scoring, write/read gates, and CRM-integrated governance.

Organizational Memory tracking 120+ business signals across enterprise tools. First AI agent with deep company context.

1,000+ enterprise integrations with HIPAA/GDPR/SOC2 compliance. Multi-tenant context management.

Enterprise agentic platform with multimodal indexing, retrieval, and extraction. "Context engineering" paradigm.

Enterprise conversational AI with memory and context management across channels.

Enterprise memory systems for AI infrastructure. GPU-optimized memory operations.

Coarse-to-fine prompt compression achieving 20x compression with 1.5% performance loss. Budget controller + token-level iteration.

Long-context compression with 4x ratio + 21.4% improvement. Document reordering + dynamic ratios.

Small BERT-level encoder via GPT-4 distillation. 3-6x faster, task-agnostic.

Semantic token importance ranking for attention-aware compression (part of Semantic Kernel project).

Extractive + abstractive document compression for RAG. 6% compression with minimal loss.

Generalized extreme compression — compress up to 500 tokens into 1 special token.

Simple local LLM inference framework. Easiest setup for running models locally.

Open-source LLM platform with integrated RAG, agents, and memory.

User-friendly AI interface for Ollama and OpenAI API. 282M+ downloads.

Lightweight local CPU inference — no GPU required.

Chat with private documents, fully on-premise. Privacy-focused.

All-in-one AI platform with RAG-centric memory and workspace management.

Benchmark for long-term interactive memory. 500 questions testing 5 core abilities across up to 1.5M tokens.

Long Conversational Memory benchmark for evaluating agent memory across extended dialogues.

75,336 Q&A pairs — 150x larger than LongMemEval. Enterprise-scale conversational memory evaluation.

Unified benchmark integrating LoCoMo, LongMemEval, and ConvoMem.

ICLR 2026 memory evaluation via incremental multi-turn interactions. Temporal memory evaluation.

Developer-focused memory benchmark for coding agent evaluation (no public repo confirmed).

Meta-prompting and context engineering system for Claude. Systematic approach to agent context.

Collection of agent skills specifically designed for context engineering workflows.

Introduction to context engineering for AI coding. Tutorial and reference.

Art and science of filling context windows — frameworks and patterns.

Proactive context-aware AI partner from ByteDance's Volcano Engine.

Adaptive web scraping framework with MCP integration for agent data ingestion.

Web data API converting websites to LLM-ready markdown. Agent-friendly output.

Python scraper using LLM and graph logic for structured extraction.

LLM-friendly web crawler and scraper. Clean markdown output for agent consumption.

Simple yet powerful long-term memory manager for conversational AI.

User profile-based long-term memory for AI chatbots. Adaptive profiling.

Conversational history management with Spring AI framework.