Memento-Teams/Memento-Skills

Memento-Skills: Let Agents Design Agents

What it solves

Memento-Skills provides a framework for AI agents to develop and persist reusable capabilities. Instead of relying solely on model weights, it allows agents to author, repair, and store skills in an external memory, enabling them to improve their performance on complex tasks over time through a "Read-Execute-Reflect-Write" lifecycle.

How it works

The system uses a 4-stage ReAct architecture consisting of Intent, Planning, Execution, and Reflection. When an agent encounters a task, it routes to existing skills using a hybrid approach (keywords or LLM-based routing). If a skill fails or is missing, the agent can reflect on the execution trace to author a new skill or repair an existing one. These skills are stored in a pluggable storage system (database, file, or vector storage) and can be evolved in the background via a "Dream Daemon" that consolidates experiences into long-term memory.

Who it’s for

It is designed for developers building autonomous agents that need to be extensible and capable of self-improvement without constant fine-tuning. It is also suitable for users who want a deployable agent with a GUI, CLI, and integration into IM platforms like Feishu, DingTalk, and WeChat.

Highlights

  • Skill-Centric Runtime: A complete orchestration system including routing, execution, and reflection.
  • Self-Evolving Capabilities: Ability to automatically create and repair skills based on failed execution trajectories.
  • Multi-Platform Deployment: Native support for desktop apps (Windows/macOS), CLI, GUI, and various IM gateways.
  • Robot-like Learning: Uses a research-backed approach to grow a skill library from atomic tools into complex learned behaviors.
  • Extensive Tooling: Includes a unified tool registry with atomic tools (bash, python_repl, web) and MCP client integration.

Related

  • Project
    ReflexioAI/reflexio

    Reflexio is an open‑source platform that records LLM‑agent conversations, extracts user‑specific facts (profiles) and reusable procedures (playbooks) from corrections and expert replies, and makes this knowledge searchable for future agent turns. It offers a CLI, Python SDK, FastAPI backend, hybrid vector‑text search, multi‑LLM support, and integrations (OpenClaw, mem0). Benchmarked on the GDPVal suite, it reduces planning steps by up to 81 % and token usage by 72 % on top of a self‑learning Hermes agent.

  • Project
    TencentCloud/TencentDB-Agent-Memory

    TencentDB Agent Memory is an open‑source memory hub for LLM agents. It captures conversations, extracts reusable skills, builds a Wiki from docs, and indexes code into a CodeGraph. A central UI (Memory Hub) lets teams manage assets, set fine‑grained ACLs, and bind specific memory pieces to individual agents. Agents talk to a uniform proxy, so any framework (Claude Code, DeepSeek Harness, OpenClaw, etc.) can use the same memory without code changes. Benchmarks show a 59 % boost in persona retention. The repo provides one‑click Docker deployment, migration tools, and a roadmap for zero‑config cold starts and more integrations.

  • Project
    NousResearch/hermes-agent

    A self-improving AI agent with a built-in learning loop that creates and refines skills from experience and persists knowledge across sessions.

  • Project
    fuyuxiang/echo-agent

    Echo Agent is an open‑source, self‑hosted AI‑agent runtime that provides long‑term memory, automatic skill extraction/evolution, and pluggable tool integration. It can run interactively in a terminal, as a long‑living Gateway service with a web dashboard and API, or be embedded in a desktop application (EchoAgent Desktop). The core runtime manages the agent loop, session context, model routing, permissions, and automation channels, while the memory system recalls past facts and the evolution system turns execution traces into reusable Skills. Install via `pip install "echo-agent[all]"`, configure an LLM API key, and start with `echo-agent run` or `echo-agent gateway`.

  • Project
    agentspan-ai/agentspan

    A durable runtime for AI agents that prevents execution loss during process crashes and supports long-running, event-driven workflows with human-in-the-loop approvals.