vectorize-io/hindsight

Hindsight: Agent Memory That Learns

What it solves

Hindsight is an agent memory system designed to move beyond simple conversation history or basic RAG. It enables AI agents to learn and synthesize information over time rather than just recalling fragments of past interactions, solving the problem of long-term memory decay and inaccuracy in complex agentic workflows.

How it works

Hindsight organizes memories using biomimetic data structures divided into world facts, experiences, observations, and mental models. It operates through three primary functions:

  • Retain: Extracts key facts, entities, and temporal data from input using an LLM and stores them in a normalized format.
  • Recall: Retrieves memories using four parallel strategies—semantic (vector), keyword (BM25), graph (entity/causal links), and temporal (time range)—which are then merged and reranked.
  • Reflect: Performs deep analysis of existing memories to form new connections and synthesize understanding, allowing agents to answer complex questions that require reasoning rather than simple lookup.

Additionally, it consolidates related facts into "observations" (evidence-backed beliefs) and maintains "mental models" (standing answers to specific questions) that act as a cache of settled knowledge.

Who it’s for

Developers building AI agents, coding assistants, and enterprise AI applications that require persistent, evolving memory across sessions for users, projects, or specific agents.

Highlights

  • State-of-the-art performance: Achieves high accuracy on the LongMemEval benchmark.
  • Extensive integrations: Supports over 60 integrations, including LangChain, LlamaIndex, CrewAI, and various coding agents like Cursor and Claude Code.
  • LLM Wrapper: Allows adding memory to existing agents with minimal code changes via a wrapper for OpenAI and Anthropic SDKs.
  • Flexible Deployment: Available as a managed cloud service, a Docker container, or an embedded Python library (no server required).
  • MCP Support: Includes a built-in Model Context Protocol endpoint to expose memory operations as tools.

Written about in

Related

  • 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
    Growth-Kinetics/DiffMem

    DiffMem is a production‑ready FastAPI service that gives LLM‑powered agents a Git‑backed, Markdown‑based memory. Each user lives on an isolated Git branch; the writer agent commits conversation updates, and a retrieval agent uses plain `git`/`grep` commands to fetch the current view or dive into history. The system needs only an LLM API key, runs on a tiny Linux box (Docker or Coolify), and optionally mirrors branches to a private GitHub repo for backup. It includes consolidation tools (dedupe, redistribute, link) to keep the “now” files compact and well‑linked.

  • Project
    onyx-dot-app/onyx-foss

    Onyx is a self-hosted knowledge and context layer that connects to over 50 applications to provide AI agents with secure, relevant access to your organization's internal data.

  • 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
    memodb-io/Acontext

    Acontext is an open-source memory layer that captures AI agent learnings as human-readable Markdown skill files, making agent memory transparent, portable, and easy to debug.