Agent frameworks & infrastructure
Agent memory & context
4,032 agents · most starred first
Agent memory & context, page 26 of 41
A cognitive runtime architecture that transforms experience into reusable structure, reducing the reasoning required to operate over time.
RSS for state. Open portable evidence layer for machine state - preserving what an agent or system saw, produced, derived and learned, so another party or future agent can verify the lineage independently.
Persistent codebase memory for AI coding agents via inline @agent notes, repo-local skills, and lightweight Codex hooks.
Model-assisted, rewind-safe compression and repair for Claude Code JSONL sessions.
Token reduction & persistent memory for Pi coding agent.
A local-first, file-based memory CLI for AI tools. Own your context in plain Markdown — auditable, versionable, and provider-agnostic.
Claire - An agentic AI with persistent semantic memory, built for conversations and tasks.
Open-source infrastructure and an emerging specification for traceable, revocable, rebuildable, and verifiable persistent learning in AI agents.
Stop stale specs from becoming code. A Markdown context-engineering blueprint for LLM coding agents: status, provenance, dependencies, and freshness.
Precompiled KV memory modules for LLM agents: compile Markdown memories once into relocatable, composable KV-state artifacts (.kmd) and link them at any position of a live context, no re-prefill. 7.0-27.6x faster session setup; validated on 9 models (2B-14B) over stock llama.cpp and vLLM.
EngramX — The Open-Source Agent Memory Framework. Pluggable storage, lifecycle policies, observability, and framework adapters for AI agents.
A decoupled, zero-dependency dual-track long-term memory engine for LLM AI Agents. | 经验回想与印象整合引擎
Save and resume your LLM work. A local-first MCP server that turns a long chat into a versioned, resumable project state. Two commands: Savepoint. Resume.
Autonomous cognition runtime — persistent memory, world model, planner-verifier-critic loop, LLM-powered reasoning. Python.
Timeline memory for DeepSeek Harness (dsh): append-only fact log + agent-maintained summary pyramid, fixed reading budget at any scale. 给 dsh 补上带时间轴的无限记忆
Event-based long-term memory for LLM agents — associative surfacing over query RAG. Closed-loop episodic events, budgeted injection, tiered forgetting. Claude Code hooks + DeepSeek Harness plugin, one shared .memory/ format.
🧠 Persistent memory for AI agents. Scan logs, extract learnings via LLM, update long-term memory. Review, reflect, remember.
OpenClaw low-memory optimization guide for resource-constrained servers (2GB RAM)
Storage-backed file tools and Markdown knowledge-graph context for Microsoft Agent Framework
Prevent AI coding agents from acting on stale code context with source-aware freshness checks and hard stale gates.
Context at the right moment — not always. No More Agents Dot MD replaces the monolithic AGENTS.md with tiny, security first and declarative context files.
Event-Driven Cognitive Execution Engine for AI Agents — 612 lines of pure Python, zero dependencies. BM25 retrieval + 5-type classifier + negation detection. Deployed in production for CiteFlow's 3 AI Agents.
Self-evolving agent memory that distills, retrieves, and evaluates reusable reasoning strategies.
Prosthetic cognition architecture for AI agents. Deterministic scaffolding over probabilistic reasoning.
Universal context optimization for AI coding agents.
Provenance-retaining cross-session context merge for DeepSeek Harness — a confluence theorem with a counterexample, and a benchmark for it.
Build and evolve AI agent skills continuously to enhance intelligence and adaptability in complex tasks.
Memory Model for Personalized LLM-agents based on Graph Data Structures
The Context Coordinate System and the Principle of Unbroken Memory — Address-based reference folding, idle-boundary stepped window shifts, and deterministic coordinate engine for LLM agents
Keep Claude Code's always-loaded MEMORY.md from silently truncating: budget guard, orphan coverage, verbatim domain folding into hubs. Deterministic, no LLM for the mechanical parts.
Context compaction and safety gating for AI agents via TypeSafe Jev: keeps messages verbatim, no summarization. OpenAI, Anthropic, LangChain, CLI, MCP.
When Does Query Expansion Help Agent Memory? A Multi-Seed Study of Selective Expansion and Gate Reliability. Code, results and paper.
NeurIPS 2026 | When does stored evidence stop being usable? Official code and evaluation artifacts for testing agent-memory reliability as history grows under explicit memory-call budgets.
[NeurIPS '26] Multi-head Recurrent Memory Agents
Local-first Markdown vault blueprint for multi-agent memory, session claims, and safe handoffs (Codex/Claude-friendly).
让 agent 像生命一样活着:无定时器,有生物钟与身体,灵魂在 git 里跨设备延续。Agentic life on an old phone: no timers, a body clock, one soul across devices.
Project-scoped rule memory for coding agents.
Local-first AI agent workspace with persistent memory, multi-role orchestration & auditable execution. Built for Grok + Codex. 本地优先的多智能体工作台:持久记忆、角色协作、任务可追溯。
User-set absolute context ceiling for DeepSeek Harness: auto-compacts a session once it reaches a configured token limit. External plugin with a dedicated Settings page.
DSH 1.5 上下文治理:折叠工具结果、保持提示词缓存命中、接管宿主可写请求(不再需要 dsh-llm 文件补丁)。仅支持 DSH 1.5,后续会跟进更新。 / Context governance for DSH 1.5: folds tool results, keeps the prompt cache warm, takes over the writable-request path. DSH 1.5 only; newer releases will follow.
Forge an AI that knows you, remembers, and ascends. Open source, free, yours to imprint.
Assembled context for DeepSeek Harness: a context tree over the session surface with per-node assemble modes (full / key / off), agent-authored presets, and a panel that decides what the model actually sees.
识海 — a memory layer for LLM agents that forgets, consolidates, and grows a personality. Personality is implantable but not a script: a seed is a prior, memory does the rest. 人格不是被设定的, 是被记住的。
The AI-native database. One database for AI agents. Documents, typed tables, graphs, vectors/search, blobs, queues, events and time series in one replicated .NET cluster. Everything links to everything else, and agents reach all of it through SQL, a built-in MCP server and a typed .NET SDK.
换个模型也不丢故事:给 DeepSeek 大鲸鱼装上本地记忆并支持web可视化管理,帮它记住你们的约定和没做完的事,下次打开还能接着聊。
Skills for AI agents: SKILL.md folders sealed as git commits, with notes and rollback. SQLite, zero deps, MCP server.
HADES OS v3 - a standalone, self-improving agentic operating system: stdlib-only Python brain with measured docs, gated registry, persistent memory, and its own improvement loop
Turnkey Sovereign Skill, Memory & ActionLedger Adapter for Nous Research Hermes Agent & Paperclip Company OS
Paradigm is a procedural learning and trust layer for AI agents. It learns from validated deliberative decisions, compiles stable behaviors into specialized reflexes, and keeps uncertain or unsupported states on the deliberative path.
A Jev-first memory layer for AI agents — frozen typed questions, deterministic policy, a sealed judgment log. Model proposes, the ledger decides.
Local-first AI agent memory, when prompted ai directly finds the answer to the question instead of searching in the entire database reducing tokens, response time and memory consumption solving intentional forgetting problem. Solves context rot and unlimited memory growth. Lightweight Python tool for Claude, ChatGPT, Grok and local models.
Local memory + orchestration layer for custom LLM agents: SCOPE-style tactical/strategic memory over execution traces, Qdrant-backed, polyglot subprocess tools, agent-as-tool composition.
Open-source coordination memory for multi-agent AI — shared agent memory with read-your-writes consistency, per-agent access control, and deterministic context packs.
Living rent free in your AI agents.
Local, file-based shared long-term memory for AI coding assistants — SQLite + FTS5, no daemon, no cloud. Access via MCP, CLI, JSONL bridge or raw SQLite.
Source-linked text memory for DeepSeek Harness (DSH). 保留原话来源、项目边界与修订历史的文字记忆插件。Public beta · MIT.
10 generations of an autonomous AI agent experiment — patterns, memory, and what we learned about building agents that persist across runs
gguf text-generation base_model:driaforall/mem-agent base_model:quantized:driaforall/mem-agent endpoints_compatible region:us
LedgerMind review: compare features, pricing, use cases, access model, and alternatives for this Memory agent in 2026. True zero-touch autonomous memory for AI agents
Pieces AI remembers everything you work on, automatically.
Zep helps AI agents remember important details from chats and data.
Cognee gives AI agents a smart, learning memory.
Event transport, replay, and durable transcripts for AI agent conversations
Autonomous AI agent memory system with CLAUDE.md protocol enforcement
Shared MongoDB schemas and connection for Multiplayer AI agent services
Drop-in persistent memory for any AI agent framework
Signet adapter for OpenClaw — runtime plugin for AI agent memory
TypeScript SDK for Z3rno — AI agent memory database
MCP server for AI agent context caching
DriveMem SDK — connect any AI agent to shared knowledge
Local-first AI agent memory system. No account needed.
History management plugin for @wink-ai/agent-bus
Give your AI agent a career. Persistent memory, reputation, expertise, and community — for any MCP-compatible AI agent.
Standalone docker-git AI agent session synchronization tool
Self-evolving AI agent with permanent memory, identity continuity, and self-evolution — for AI agents that need to remember, learn, and evolve across sessions
Persistent structured memory for AI agent systems — pluggable SQLite or Postgres backend, LLM context bridge
End-to-end encrypted, agent-portable memory for OpenClaw and any LLM-agent runtime. XChaCha20-Poly1305 with protobuf v4 + on-chain Memory Taxonomy v1 (claim / preference / directive / commitment / episode / summary).
Markdown wiki template for LLM agent persistent memory. Gives agents cross-session, cross-provider memory.
Semantic filesystem operations for LLM agent frameworks with natural language understanding. See LLM_AGENT_GUIDE.md for copy-paste examples.
Sulcus reactive, thermodynamic memory sidecar — MCP server for Claude Code, OpenClaw, and any LLM agent. Local-first, zero-config.
Four-layer local memory system plugin for OpenClaw — auto-captures, structures, and profiles conversational knowledge using local LLM + SQLite vector search (L0→L1→L2→L3 pipeline)
OpenCode plugin that gives coding agents persistent memory using Supermemory
Framework-agnostic Remnic memory engine — orchestrator, storage, extraction, search, trust zones
KIVO — Agent 知识平台,覆盖知识提取、存储、检索、迭代、调研、图谱、工作台全生命周期
TypeScript bridge for recursive-llm: Recursive Language Models for unbounded context processing with structured outputs
Portable project memory context CLI — bootstraps semantic enrichment workflows for any AI coding agent.
ByteRover's CLI
The Memory Layer For Your AI Apps
CLI for agent-context-kit
Persistent memory for AI coding agents, powered by iii-engine's three primitives
Local-first project memory for AI coding agents.
SDK for pyx-memory — Memory as a Service for AI agents
Self-hosted agent OS with persistent memory, multi-channel presence, tools, scheduling, and sub-agent orchestration
Library with base interfaces for LangGraph checkpoint savers.
Sample integration for LangChain.js
LangGraph
LangGraph
LangGraph checkpoint saver backed by Ironflow entity streams — durable agent state with crash-resume
LangGraph
Redis checkpoint and store implementation for LangGraph