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
alexisfox7/PRO-LONG
PRO-LONG is the reference implementation of a paper on programmatic memory for long-horizon agent tasks. The idea is deliberately minimal: instead of a specialised retrieval system, a subagent hierarchy or an elaborate memory module, the harness appends every observation, action and outcome to a single structured text log, and the agent reads that log with ordinary tools, grep and Python code, whenever it needs history. The system prompt is about thirty lines. Applied to the public ARC-AGI-3 game set, this addition raised the score of the same underlying coding agents by an average of 18 percentage points over running them without the log, matched or beat purpose-built harnesses while billing between four and six times fewer tokens, and reached a reported best-of-two score of 97.4 percent. The code supports two backends, the OpenAI Codex command line tool and the Claude Code command line tool, each running inside a Docker sandbox that mounts only the game workspace and has no network access apart from a proxy to the model API. A swarm command line entry point runs whole game suites or named games with configurable action limits, model and reasoning effort. Ablation conditions are exposed as flags so the memory claim can be tested directly: a truncated log window, no log at all with the current board placed in the prompt, and a stateless variant where the workspace is wiped between calls but the log survives. The repository also publishes official scorecards that can be verified on the ARC Prize site, plus release logs containing game logs, agent transcripts and workspaces. It suits researchers studying agent memory, context management and long-horizon reasoning.
https://github.com/alexisfox7/PRO-LONG
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.
modelscope/AgentEvolver
Self-evolving agent training framework from ModelScope that combines automatic task generation, experience-guided exploration and attribution-based credit assignment into one reinforcement learning stack.