thedotmack/claude-mem
Claude-Mem is a plugin for Claude Code that captures, compresses, and preserves coding session context to enable persistent memory and continuity across sessions.
Awesome AI Agents › Continual Learning Agents
modelscope/AgentEvolver
AgentEvolver is an end-to-end training framework for agents that improve themselves, built around three self-evolving mechanisms that span the path from environment to policy. Self-questioning has the agent explore its environment and generate diverse tasks on its own, which removes the need to hand-build training datasets. Self-navigating summarizes experience gathered across earlier tasks and reuses it to guide later rollouts, raising the quality of exploration. Self-attributing analyses long trajectories to work out how much each intermediate step contributed to the outcome, giving finer-grained signals for policy optimization than a single end reward. The system uses a service-oriented dataflow architecture in which environment sandboxes, language models and experience management run as modular services. Standardized interfaces connect external environments and tool APIs, a context manager handles multi-turn interaction and long contexts, and decoupled components are meant to make customization and algorithm upgrades straightforward. Reported results on the AppWorld and BFCL v3 benchmarks show large gains over the untuned Qwen2.5 baselines at both 7B and 14B parameter sizes, with each additional mechanism contributing, and the tables list avg@8 and best@8 scores for every configuration. Getting started requires conda and a CUDA toolkit, an install script, setup of an environment service such as AppWorld, and optionally ReMe for experience management; a launcher script starts the environment, the log dashboard and the training run together from a YAML configuration. The repository also hosts a Game Arena that extends the framework into multi-agent social reasoning games such as Avalon and Diplomacy for interaction, evaluation and training, plus research branches for a sequence-level agentic reinforcement learning method and an extended self-questioning method. Written in Python 3.11 or newer and licensed under Apache 2.0.
https://github.com/modelscope/AgentEvolver
Also in Continual Learning Agents
vectorize-io/hindsight
Hindsight is an agent memory system designed to create smarter, learning AI agents, addressing limitations of traditional memory techniques like RAG and knowledge graphs.
langchain-ai/langchainjs
LangChain.js is a TypeScript framework for building context-aware, reasoning applications powered by large language models, offering modular components, chains, agents, and tools for production and deployment.
aden-hive/hive
Hive is an outcome-driven agent development framework designed for evolving agents with AI evaluation, human-in-the-loop, and self-improvement capabilities.
aiming-lab/MetaClaw
Proxy-based layer that turns everyday conversations with a personal agent into learning signals, injecting skills per turn and running scheduled reinforcement learning updates without a local GPU.
kayba-ai/agentic-context-engine
Agentic Context Engine (ACE) is a framework that enables AI agents to autonomously learn from their successes and failures, continuously improving their performance across various tasks without additional training data or fine-tuning.
AgentDock/AgentDock
AgentDock is an open-source framework for building sophisticated AI agents with configurable determinism, enabling reliable and creative AI applications through a node-based architecture and multi-stage workflows.
ghostwright/phantom
Self-hosted persistent AI co-worker built on the Claude Agent SDK that runs on its own machine with long-term memory, self-evolution, an MCP server and Slack, Telegram, email and web channels.