Agent frameworks & infrastructure
Agent memory & context
4,032 agents · most starred first
Agent memory & context, page 21 of 41
local-first persistent memory for Claude Desktop, enabling private hybrid RAG search across past conversations without API calls
Nyx — 夜之感知层。Veil Bloom Filter + Mist Phantom 寻回失落记忆,Thalamus 路由 + Hypnos 梦境的第三子系统。
Context compression for AI agents — cut token usage costs with retrievable CCR compression. Rust core + Python API, Claude Code & Codex plugin, MCP server. PyPI: furl-ctx
Obsidian vault template that gives AI agents persistent, shared memory — Claude Code, Codex, and Cursor read and write the same second brain. AGENTS.md rules, per-project memory, human-gated principles.
An asynchronous memory engine for enterprise AI agents, statistically minimizing hallucinations via dual-LLM consensus
A multi-agent AI system with isolated agent memory, deterministic confidentiality guards, transactional context compaction, provider adapters, and replayable observability.
File-first memory runtime for AI agents — versioned, portable, and always there when the next session starts.
Codex-native Lossless Context Management
AI-native long-term memory for Hermes Agent with AES-256-GCM encryption at rest, dense HNSW ANN, sparse and exact-text recall, metadata/range filters, and MinHash dedup, backed by an embedded or multi-client server columnar SQL database with 35 language bindings.
Universal Shared MCP Memory Server - Multi-tool compatible memory management for AI assistants across Claude Desktop, Cursor, Windsurf, Cline, and more.
Open protocol for AI agents to use Obsidian vaults as persistent memory with SHA256-based incremental indexing.
Shared memory MCP server for multi-agent projects. Agents (Claude, Cursor, Codex, AntiGravity) write to a local `.agent-mem/` folder so the next agent picks up where the last one died.
AI employees with self-improving context that actually get stuff done
Local-first shared memory and task coordination for AI coding agents. One Go binary, MCP server, markdown files you own. Hooks for Claude Code and Codex CLI (and their desktop apps).
Persistent, shared memory for AI agents - so they stop forgetting and stop colliding. Markdown-native, on-prem, auditable. Speaks MCP.
Memory for your coding agent — markdown project graph template: product, plans, features, bugs, specs, architecture, releases
Local fine-tuning for your own work. A provider-neutral kit that turns project experience into reusable criteria for AI agents — distilled at a threshold you control, pruned when it grows, and portable between models.
The interaction primitive.
Fix context rot & pollution in long AI agent sessions. Subagents run in isolated processes — main agent stays read-only. Context isolation via process isolation.
The context layer for AI agents — open, structured, searchable knowledge protocol (Domain → Concept → Link). Skills-first, no MCP server needed.
Amazon DynamoDB storage backend for the Strands Agents SDK (TypeScript and Python): sessions, agent memory, semantic search over DynamoDB vector indexes, and S3 offload behind one Storage interface.
Computer context layer for agents
Cortex is a decentralized sovereign elastic memory layer built on the Sui stack that lets people, agents, and applications store, share, and act on durable encrypted intelligence.
Local-first long-term memory for AI agents - plain Markdown vault + thin MCP server. CJK-aware BM25, token-efficient by design.
One workbench for AI coding agents—run locally or remotely, review every task, and keep memory and Markdown knowledge.
Self-evolving Hypergraph Memory — 自演化超图记忆引擎: τ 衰减遗忘 · Hebbian 关联强化 · 离线梦境整合 · 多模态证据检索 · 谓词论元与类型锚定抽取 · 确定性时间层 · 会话作用域隔离 · 本体驱动组织 (FastAPI + OverGraph + FAISS, 全开源)
The navigation of AI Agent.
Proactive context synthesis for AI coding agents. Build the right context before the first token.
A portable, self-hosted memory plane for AI agents — one core across API, MCP, CLI, and UI.
Pass your project, not your context.
Willow 2.0 — local-first memory, MCP tools, and fleet coordination for AI agents. Postgres KB, Ollama default, Grove LAN remote.
Persistent memory graph for AI coding agents — semantic search, knowledge graph, and time-based decay over MCP
为 Claude Code 等 Agent 打造的分层记忆管理系统,让 AI 跨会话、跨项目、跨机器持久化并共享项目与个人知识。
Benchmarking Cognitive Traps in LLM Memory Use
Open-source protocol for managing AI resources with dependency/package semantics across agents.
A Yogācāra (唯识) self-model plugin for DeepSeek Harness: eight consciousnesses, the 51 mental factors, a perfumed seed store, and a measurable self — written back into the agent's own prompt.
MCP-native knowledge-graph memory for AI agents
Synchronize your AI coding agents' memory so you can stop in one agent and pick up in another
Project and session organization for DeepSeek Harness: multi-folder projects, searchable chats, attention summaries, and native folder picking.
A memory hub and curated skill set for Claude Code — persistent memory that loads from any directory, plus a guided local setup.
Self-hosted, self-learning, provider-agnostic personal agent kernel. Local-first memory in one SQLite database, project-scoped, reachable from any tool over MCP or a CLI.
Makes the branch the building block of AI conversation management — parallel workflows, continuous and mergeable conversation memory
Sleep-time memory consolidation for Claude Code agents, with a cross-project shared-consciousness layer.
Project Memory gives coding agents a single, trustworthy memory for a software repository — so they stop re-learning the same facts and stop writing conflicting "memory" files.
Pi-native implementation of SKILL.state for scalable long-horizon agent execution
Rate limit? Switching coding agents? Stele keeps write-ahead task state so work resumes in any harness with nothing to re-explain.
Grounded perceptual memory for personalized agents: VLM grounding, specialist identity encoders, and a training-free in-model attention bank. Paper: arXiv:2608.28609
类脑联想记忆库:SQLite + bge-m3 + 扩散激活的单机记忆引擎。纯向量回答'哪条最像',联想记忆回答'这整件事是什么'。Brain-inspired associative memory engine (spreading activation, online learning, flexible forgetting).
Persistent knowledge and validation-gated skill evolution for coding agents, based on arXiv:2608.27454.
Cross-tool persistent memory for AI agents — local, private, no API key required
Local-first shared memory for AI agents — zero-dependency, governed writes, citations, MCP. Source behind the ihow-memory npm package.
A self-improving ML agent that compounds capability across projects through knowledge crystallization, reusable skills and shared ML/MLOps infrastructure.
Local-first persistent memory for AI agents: Chinese-first recall, forgetting-aware ranking, conflict candidates, SDK + CLI.
BrainAgent — brain-inspired plugin for DeepSeek Harness (dsh): a pipeline of heuristic filters, statistical memory stores and context injections; reward-ledger + UCB1 bandit learning loop (RL-lite). 915 tests. Free noncommercial use.
The memory is yours. The model is a guest. Local-first memory and retrieval layer for AI agents: RAG with no model in the path, abstention as a verdict, MCP server, evaluation included. Rust, Apache 2.0.
Context Management Layer for Agentic AI. Graph-based knowledge context, SQLite index, advanced FTS5 lexical/BM25 search and cross-session live memory.
ENGRAM — an epistemic memory substrate for AI agents. Public release snapshots (one squashed commit per version).
Long-session handoff for Hermes Agent: measure how close a chat is to the token limit that ends it, export the whole session, and continue in a fresh chat that reads the parent transcript in full. Recovers context that compaction already removed.
Portable memory layer for coding agents — persistent user & project context across sessions, installed into Claude Code.
The local-first brain for coding agents: project memory + workflow intelligence + lossless ~50% context savings. MCP server + LLM proxy, zero cloud.
TMCRA Core — local-first, scope-isolated long-term memory runtime for AI agents.
Git-backed, conflict-safe memory authority for AI agents via MCP (Model Context Protocol).
A context layer for coding agents: notes, RFCs, tasks and tracked time as one graph of plain markdown. The running app hosts an MCP server, so Claude Code and other agents read what is ready, follow the reasoning behind it, and log their own work.
Take your AI's memory anywhere - a local-first, portable, git-like memory layer for AI coding agents (MCP-native, $0/offline by default).
🧠 Mnema — Long-term memory for AI via MCP × Vector DB. Pluggable backends, hybrid search, Auto Dream consolidation, CLI + REST API + SDK.
Use ChatGPT with a private local memory
Local-first persistent memory plugin for opencode: three-layer memory (long-term knowledge / episodic summary / working memory) with SQLite + Chinese-aware retrieval (RAG), fully local.
Portable, local-first memory layer for AI coding agents. One .agent/ folder shared across Claude Code, Cursor and Cline over MCP. Plain files, no vector DB.
A governed, local-first memory authority for AI agents.
Local-first universal session store for AI agents, powered by TOAST.
3-tier hybrid memory plugin for OpenCode — compressed memory, atomic SQLite archive, session isolation, endless sessions via soft compaction
Local markdown memory & cross-agent context engine for coding agents (MCP).
Persistent intention infrastructure — goals, decisions, commitments, and strategic reasoning for AI agents
Admin dashboard and CLI for Statewave — manage episodes, memories, and subjects.
Provenance and decay for AI agent memory. Audits CLAUDE.md, AGENTS.md, .cursorrules, GEMINI.md and 18+ agent memory files against the live repo: typed claims, cited commits, half-life decay, token meter, approve-only fixes. Works with any agent.
The context layer for AI agents — open, structured, searchable knowledge protocol (Domain → Concept → Link). Skills-first, no MCP server needed.
The trust baseline for multi-agent memory — event-sourced, governed, federated. 事件溯源 · 治理管线 · 联邦同步
Manage AI conversation lifecycle: archive chats to named files, free up context, recall on demand. OpenClaw skill.
Persistent, federated meta-harness for AI agents. Continue, delegate, fork and merge work across Claude Code, Codex, Cursor, Kimi, OpenClaw and more.
Local, cross-AI memory for coding agents: Claude Code, Codex, Gemini, Antigravity, Cursor, Kimi & more share one private corpus. Hybrid search, auto-injected context, a shared skill library, a work board so sessions don't collide, and homework for ideas parked mid-task. No-API-key summarizers; runs fully local via Ollama.
Project context and memory for coding agents.
One memory for every agent — shared cross-agent memory hub via MCP & REST (TS + Python + Chrome extension) / 让每个 AI Agent 共享同一段记忆
The complete operating system of an AI agent running a real business. Soul, skills, infrastructure, everything.
Sistema de Memoria Dual MCP con Búsqueda Híbrida (SSC/BIRE) — PostgreSQL + pgvector + pg_trgm + embeddings locales (Ollama)
AI Agent 的项目记忆层 — 知识图谱 + MCP,让 Claude Code / Codex / WorkBuddy 拥有项目长期记忆
Switch agents. Not context. Context bridge across Claude Code, Codex, Grok and Antigravity sessions.
Forge a second brain your AI agents can read — a CLI that generates and validates AI-readable memory vaults. A Sunsato product.
Turn agent work into governed, reusable judgment assets — a local-first personal knowledge operating system.
Progressive disclosure retrieval for long-horizon AI agents over personal knowledge bases. Three-tier model: hot cache → FTS5 index → full read. Local-first, zero dependencies.
A memory for your AI coding assistant. It keeps what you tell it and hands the right note back at the moment it matters. Runs entirely on your own machine.
Never lose your AI context again. Open-source memory that follows your projects across Claude Code, Codex, Gemini CLI, Cursor and every computer.
Local-first long-term memory for AI agents. MCP, hooks, and handoff across Cursor, Claude, and other IDEs — no cloud, no API keys.
Embedded Rust storage for agent checkpoints, hybrid memory retrieval and recoverable local work. Native Node.js and Python SDKs.
Local-first memory for you and your AI. One SQLite vault on your computer, shared across local and cloud AI tools through MCP or the CLI.
The repository is the project memory: plain files, versioned by Git, for humans and coding agents. One layer was missing — the why.
Dreaming plugin for OpenClaw memory-lancedb — Light/REM/Deep sleep phases with narrative generation and memory promotion.
SQLite of agent memory — single-file, embeddable, zero-config. Episodic + semantic memory, hybrid BM25+vector recall, bitemporal facts, spreading-activation, working memory, async API, LLM compression, Ebbinghaus decay, GDPR forget, multi-agent, MCP-native.
Shared memory for AI coding agents: summarizes sessions into plain-Markdown notes in your own Obsidian-compatible vault
Persistent memory engine for AI coding agents — preserve project context, retrieve relevant knowledge, and maintain consistency across sessions using MCP.
Long-term conversational memory and a context compiler for multi-agent character chat. Runs offline; plug in any OpenAI-compatible model.