Agent frameworks & infrastructure
Agent memory & context
4,032 agents · most starred first
Agent memory & context, page 9 of 41
Look Back to Reason Forward: Revisitable Memory for Long-Context LLM Agents
源码级逆向工程(Hermes、OpenClaw、nanobot、NullClaw、OpenFang)记忆系统 —— 架构图、数据模型、检索管线、复刻指南。Source-Level Reverse Engineering of Memory Systems (OpenClaw, nanobot, NullClaw, OpenFang) — Architecture, Data Model, Retrieval Pipeline, Replication Guide
An open-source AI chat client with structured memory and persona engineering.
The memory substrate for AI agents and developer tools. Git-versioned, file-native, MCP-first.
Memory system for AI agents. #1 on LongMemEval — 96.2% (481/500). Beats every published system including Chronos, Mastra, Supermemory, and Emergence. Built solo in 16 days for $1,000.
Persistent memory and operational continuity for AI coding agents — append-only event ledger (AgentBook), rolling summaries, self-model, belief knowledge, and turn-1 frontpage injection for deterministic cold-start.
Memory database for AI agents — stores typed memories (facts, preferences, events, entities, relations) with provenance, confidence, and bi-temporal validity, and answers with fused semantic + lexical + entity-graph + temporal retrieval. One Rust core, one binary wire protocol. Apache-2.0.
Event-sourced runtime for reconstructable AI identity, memory, reflection, commitments, and auditable cognitive continuity.
Know when your AI is about to forget — browser extension that monitors context window usage
Adam Framework for OpenClaw — 5-layer persistent memory and identity architecture for AI agents. Production-validated over 353+ sessions. First documented case of emergent values in persistent AI, quantum-verified on IBM hardware.
A Python library for long-term memory in language models. Improve conversational scenarios and create autonomous learning agents with enhanced context.
Give AI agents the ability to edit their own context window
Wavelet-based multi-resolution context management for LLMs via MCP
Local-first memory layer for CLI agents. Indexes Claude Code, Codex CLI, Cursor, Antigravity, and Pi sessions into SQLite + FTS5, searchable from a CLI, TUI, or MCP server.
Self-improvement loop for Hermes Agent: reads its own trajectory, proposes and applies the smallest skill/memory edit (create/patch), journaled with rollback. /refine command, refine_run tool, on_session_end hook.
Learn to design an LLM wiki from first principles. Used as agent memory. Workshop with 3 exercises, open-source code, video, and slides. No coding skills required.
A Memory Engine for AI Agents,Working with Jev can bring higher decision accuracy . Nylon models a memory as a woven body of multiple strands of "threads" (facts/emotions/timing/relationships/beliefs/frequencies), with weighted edges between memory nodes forming a network graph rather than a hierarchical tree.
A curated list of awesome Letta projects, tools, tutorials, and resources for building stateful AI agents with persistent memory.
Cursor uses AI to edit code — we use AI to edit AI's context. 🪆 Context map + compression + version control for LLM context windows.
A collection of pre-build wrappers over common RAG systems like ChromaDB, Weaviate, Pinecone, and othersz!
Persistent memory graph for AI agents. Facts, decisions, entities, and relationships that survive across sessions, tools, and providers. MCP server — works with Claude, Cursor, ChatGPT, and any MCP client.
Local-first context ingestion and retrieval for AI tools. SQLite + embeddings + MCP server for Cursor & Claude.
A persistent, multi-agent Model Context Protocol (MCP) server. Continuum gives AI coding agents — Claude Code, OpenCode, Codex, Gemini CLI — a shared, live view of a codebase and a memory that survives across agents and sessions.
Turn your files into a memory filesystem for AI agents.
Self-evolving context for any coding agent.
AI Runtime Kernel: a context operating system for AI agents. Eliminates tool bloat, loads only what’s needed, and gives LLMs their reasoning space back.
Persistent project memory for AI coding agents — isolated workspaces per project, web dashboard to review what your agent remembers, team collaboration. Free to use.
An experimental framework for autonomous, self-evolving AI agents living on a server with cross-session state and persistent memory.
Session memory for Claude Code. Auto-captures every session into a plain Obsidian markdown vault. Zero config. MIT. macOS + Linux.
Open-source bitemporal knowledge graph — long-term memory for AI agents. Hybrid retrieval, conflict-aware ingest, GDPR forget. REST + native MCP.
A Python app that shows how an agent calls tools, saves results, and recalls them after a restart.
Agent team knowledge framework with layered structure for principles, skills, insights, and experience
Memory management system to enhance AI agents with personalized, reasoned and context-aware responses
Markdown-first work-memory protocol for existing agents, with maintained knowledge, candidate notes, evals, and an example KB.
TMLR | This survey presents a comprehensive and structured synthesis of memory in LLMs and MLLMs, organizing the literature into a cohesive taxonomy comprising implicit, explicit, and agentic memory paradigms.
Agentic memory built on Postgres
An open experiment in memory for AI agents — the whole memory, always in context.
Better memory for your AI agents. (Karpathy + Graphify + PCKC)
Agent-native long-term memory that learns after every run. File-native, auditable, and integrated with Pi, Hermes, OpenCode, and OpenClaw.
Persistent code memory for AI coding agents, kept in your project on your machine. Gives Claude Code, Cursor, Cline, Hermes-Agent and any MCP client the files that matter for each prompt, learns which files you use together, and publishes every benchmark with its misses.
Semantic memory for AI agents. Local-first, MCP-native.
AI 恋爱 · 人机恋 · AI 伴侣记忆引擎 —— 让你的Ai/Agent 永不失忆
Organize your AI sessions, and carry what they know into the next one.
Open plain-text file format for AI memory. Your assistant's long-term memory as .dai files on your disk: readable by Claude, GPT, Gemini, Cursor, local models and grep (all LLM models work). MCP server + hooks for Claude Code, Claude Desktop, Cursor, Windsurf, Codex. 83% LongMemEval-S (GPT-4o), 92% (Claude Fable 5), 10x fewer tokens.
Two-engine memory stack for AI agents: Mnemosyne (reflex memory) + GBrain (knowledge brain). Installer, launcher, doctor — one command, any host.
Persistent memory for Claude Code. Your AI never starts from scratch again.
Agent Memory Playground: AI Agent Memory Design & Optimization Techniques
Local-first Model Context Protocol (MCP) memory layer for Codex CLI/Desktop, Claude Code, Gemini CLI, Qwen/DeepSeek/Ollama and agent workflows. SQLite + FTS5 compact context packs, token savings, read-only mode, no external memory server.
基于本地 RAG 与大模型人格蒸馏的多平台聊天记忆系统。通过 FTS5 检索、语料清洗与长期报告生成,将聊天记录转化为高拟真的数字分身。
Give your coding agent a persistent memory. source-of-truth keeps a living docs/ catalog of what exists, why, and what rules apply — so AI stops rebuilding, deleting, and breaking things it never knew were there. Works with Claude, Codex, Cursor, Gemini, Copilot & more.
A self-evolving knowledge accumulation & hybrid retrieval system for AI coding agents (Claude Code / Codex / etc.)
Persistent, verified memory for coding agents — so they stop re-explaining your codebase and never act on stale knowledge. Every memory is checked against your actual code; lives in your repo as plain files, shared via git. No account, no DB. Install: npx -y @kage-core/kage-graph-mcp install
A lightweight, Git-backed permanent memory for AI agents.
A minimal coding-agent harness built to measure what context management actually costs: constraint survival after compaction (8%→83%), default-FAIL verification against false completions, and KV-cache prefix design (21%→90%). Stdlib-only, real deepseek-chat runs
The operating system that AI workflows trust. Structure messy inputs. Establish identity. Track provenance. Build the durable data layer your AI workflows need to be trusted, reusable, and auditable.
Context Strategy Framework: Intent → Build → Learn workflow. Preserves context in AI-generated code.
🚀 Lightweight Python library for building production LLM applications with smart context management and automatic token optimization. Save 10-20% on API costs while fitting RAG docs, chat history, and prompts into your token budget.
Memory layer for AI Agents and OpenClaw powered by Rust
Shared memory for AI-powered engineering teams. Captures, indexes, and recalls conversations across Claude Code, Cursor, Codex, and other agents. Share team knowledge through git — no cloud required.
The complete backend for AI agents in one SQLite file : Documents, vectors, cache, queue, and cron.
Self-hosted AI agent memory server with MCP, evidence provenance, typed claims, conflict detection, embeddings, recall, PostgreSQL, and pgvector
Smart Compact — recoverable context cleanup, verified compaction, session navigation and project memory for Pi Coding Agent.
Hermes-first operational AI substrate with filesystem-grounded memory, governance, persistent routes, and recoverable agent continuity.
Give Claude Code a persistent memory — it remembers you, your projects, and your decisions as plain Markdown files you own. Fully local. Just say `start`.
Reflective memory for agents
The Cost of Remembering: filesystem memory matches long-context accuracy on LongMemEval while reading 97% fewer tokens and costing 95% less. Harness, run data, 129 agent-built memories, and paper source.
DeepSeek Harness (DSH) local-first cross-session project memory: zero bundled runtime deps, bounded cache-friendly Hot Memory, BM25 recall/cache, provenance, lifecycle and bilingual Web GUI; no embeddings or external service. / DSH 本地项目记忆插件。
CPU-side action routing, context compression, and causal memory for AI agents — matches an LLM-everything agent at 58% fewer LLM calls and ~45% lower cost. Glues busyBee-cpu, honey-comb, and rust-brain.
Cross-device DeepSeek Harness session sync: a dedicated git mirror with append-only keep-both conflict resolution (fork files, never silently overwritten), /sync command and sync_* tools
Durable, traceable project memory for AI coding agents. An evolution of Karpathy's LLM Wiki pattern built for codebases.
Local-first MCP memory server for AI agents. Zero dependencies, SQLite on your own disk, 52 MCP tools with Mem0-compatible names. MIT.
Open-source procedural memory for AI agents. Learns from what worked and what failed.
Open Memory Protocol (OMP): an open protocol for moving and sharing AI-agent memory. Hosted at the AI Disclosures Project, co-convened with IBM and Mozilla.
Cost saving Cognitive Architecture for AI Agents. Infinite Context. Perfect Recall. Zero Hallucinations.
Agentic memory for CTI in Python — STIX knowledge graphs, threat-actor alias resolution, offline-first RAG, MCP server for Claude Code and LangChain agents
AI Agent 记忆管理系统:P0/P1/P2 优先级 + 自动归档,Token 降 78%
An ergonomic, opinionated memory interface for AI agents
OpenClaw plugin for persistent AI memory
You look lonely. I can fix that.
Git-commit your AI conversations before they disappear.
Topic-based automatic memory for Claude Code — never lose context across sessions or compactions.
A high-performance, local, offline-first AI memory system built in Rust
The highest-scoring AI memory system ever benchmarked that isn't reliant on LLM reranking. And it's free & burns less tokens.
Local-first agent workflow for building private cross-project memory with explicit consent, validation, and crash-safe history.
dsh-agent-team gives DeepSeek Harness agents that don't reset: durable Members with their own memory, notes, and skills across sessions, rollovers, and restarts. You set the direction; agents coordinate through Channels and Tasks.
Second brain for coding agents. Seal the day, distill into Obsidian, hit that page tomorrow. Cursor, Codex, Claude Code, DeepSeek Harness.
Retrieve by relevance, not resemblance: filter an AI assistant's memories with TypeSafe's Jev
Save conversations, search by idea, read the surrounding context. Memory for AI agents.
A pure-Python system that automatically discovers, verifies, and applies relevant historical requirements for coding agents — no embeddings, no vector database, no LLM calls. Companion code for the Towards Data Science article on intent continuity.
Self-hosted memory service for AI agents that work in teams, written in Rust: layered L0→L3 memory, async LLM distillation, hybrid BM25 + vector recall, and department/team/member isolation. SQLite or Postgres, any OpenAI-compatible LLM.
Self-hosted memory service for AI agents that work in teams, written in Rust: layered L0→L3 memory, async LLM distillation, hybrid BM25 + vector recall, and department/team/member isolation. SQLite or Postgres, any OpenAI-compatible LLM.
Reference architecture for LLM-based applications on Google Cloud Platform with Redis Enterprise as a high-performance data layer.
A forward proxy to turn network traffic into personal memory for AI agents
User profile and long-term memory for your AI agent. Drop-in LLM proxy that gives AI model persistent memory and structured user understanding.
Tocket: a file based project notebook that many AI agents read and write together
The system of record for AI agents
Memory for AI agents that never leaves your device. Local-first, end-to-end-encrypted, zero-telemetry memory engine (Rust, MCP) — gives Claude & any MCP client persistent cross-session memory.
AI agent harness tool allowing it to wipe its own session context and re-run the first message
Give your AI a hippocampus. A unified, local-first memory operating system written in pure Rust