Goldentrii/AgentRecall-X

Correction-first persistent memory for AI agents. MCP server + SDK + CLI. Compounds across sessions.

What it solves

AgentRecall provides a persistent memory system for AI agents that specifically focuses on learning from human corrections. While most memory tools focus on retrieving information, AgentRecall tracks whether a correction actually changes the agent's behavior across different sessions, preventing the agent from repeating the same mistakes.

How it works

The system implements a governed corrections ledger and a measurement harness to track the lifecycle of a correction from capture to behavioral change. It organizes memory into five distinct layers: Episodic (session journals), Semantic (topic-clustered facts), Procedural (IF-THEN rules), Narrative (project phases), and Correction (behavioral calibration).

It integrates via an MCP server, SDK, or CLI, and uses a "session loop" (start, save, recall, reflect) to ensure memory is loaded and consolidated. For enhanced retrieval, it offers an optional 100% local ONNX embedding model for semantic search without cloud dependency.

Who it’s for

Developers using AI agents (specifically Claude Code, Cursor, Windsurf, and VS Code) who want their agents to remember specific preferences, project-specific rules, and corrections to avoid repetitive errors.

Highlights

  • Behavioral Measurement: Tracks "heed rates" and "phantom gradient steps" to quantify if an agent actually stops making a mistake after being corrected.
  • Five-Layer Memory Architecture: Maps cognitive psychology to a filesystem-based memory store (Episodic, Semantic, Procedural, Narrative, and Correction).
  • Local-First Privacy: Memory is stored as local markdown files, and optional semantic embeddings run locally via ONNX.
  • Autonomous Consolidation: An optional "Dreaming" mode that runs nightly to mine patterns, decay low-signal information, and promote repeated corrections to cross-project awareness.
  • War Room Dashboard: A local-first visual interface to manage and review the agent's learned memory.

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
    beilusaiying/always-accompany

    A multi-purpose AI agent framework focused on context and attention mechanisms, providing layered memory and dynamic retrieval for long-term companionship, coding, and work.

  • Project
    atomicstrata/atomicmemory

    AtomicMemory is an open‑source, inspection‑ready memory service for LLM agents. It offers correction‑aware updates, model‑agnostic embeddings, and both managed‑cloud and self‑hosted Docker deployments. The project ships a Rust Core server, a TypeScript SDK, a CLI (`am`), adapters for LangChain, LangGraph, Vercel AI, etc., and plugins for various agent hosts. Quick start is a one‑line installer that creates a cloud profile; a `--local` flag runs the Core container on your machine. Memory can be ingested and queried via the CLI or programmatically via the SDK, and agents can be linked through an MCP server. Licensed under Apache 2.0.

  • Project
    mnemox-ai/tradememory-protocol

    TradeMemory Protocol is an open‑source Python library that records an AI trader’s own past trades, scores them with a five‑factor Outcome‑Weighted Memory model, and can block or reorder future orders before they reach a broker. It syncs fills from venues like Hyperliquid, Alpaca, MT5, and Binance, provides a local MCP server (and an optional proxy brake) for pre‑flight risk checks, and maintains a tamper‑evident SHA‑256‑chained audit log.

  • Project
    Memento-Teams/Memento-Skills

    A skill-centric agent runtime that allows AI agents to design, execute, and persist reusable skills to improve their performance on complex tasks through a reflection-based learning loop.