AIエージェント
hindsight
Agent memory that learns — so your agents get smarter over time.
★ 47.0k スターMIT
ガイドを見る →honchoは、plastic-labsが公開しているオープンソースのAIエージェントプロジェクトです（GitHub: plastic-labs/honcho）。GitHubスターは7,536、フォークは928で、AGPL-3.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月8日です。
Honcho — Plastic Labs' memory library for stateful agents that learn about their users over time.
Decide managed versus self-hosted, and model your world as peers and sessions — typically your users and your agent as peers exchanging messages.
Clone plastic-labs/honcho from GitHub (or use the install path in the run steps below) and skim its README for the parts relevant to your goal.
Install the SDK for your language; for self-hosting, clone the repo and start it with Docker Compose, then point the SDK at your local base URL.
Store a few messages on a session, let background reasoning run, then query Honcho for what it knows about that peer and inject the context into a prompt.
Store messages, let asynchronous background reasoning build understanding, query it, and inject prompt-ready context into any model.
Memory is organised around peers (users, agents) and their sessions, rather than a flat chat log.
Use the managed api.honcho.dev service, or self-host from source with Docker Compose or a local dev setup.
pip install honcho-ai, or npm install @honcho-ai/sdk, with equivalent workflows in both languages.
Documented integrations add Honcho memory to Claude Code, Codex and Cursor via dedicated packages and install scripts.
Aims at continual, cross-session memory and context engineering rather than one-conversation chat history.
Get the source from the official GitHub repository for honcho.
git clone https://github.com/plastic-labs/honcho.git
cd honchoInstall the Python SDK as the README quickstart shows (TypeScript: npm install @honcho-ai/sdk).
pip install honcho-aiClone the repo and start the full service locally with Docker Compose, then set HONCHO_URL or base_url to your instance.
git clone https://github.com/plastic-labs/honcho.git
cd honcho
docker compose up # see README self-hosting sectionplastic-labs (Plastic Labs), through the plastic-labs/honcho repository, with docs at honcho.dev/docs.
Yes — the project is open source under the AGPL-3.0 license as recorded for this listing. See the license terms for any conditions on redistribution.
The managed service uses api.honcho.dev (account/API key). Self-hosting is free under AGPL-3.0 and runs on your own infrastructure; reasoning uses models you configure there.
A database stores raw history; you still build recall, summarisation and context assembly. Honcho adds the reasoning layer — it processes messages in the background so queries return understanding and prompt-ready context, not just logs.
Clone the repository (or use the documented installer), follow the install and configure steps on this page, and run the smallest real task first — store a few messages on a session, let background reasoning run, then query Honcho for what it knows about that peer and inject the context into a prompt.
Agent memory that learns — so your agents get smarter over time.
Stateful AI agents with memory that learns and improves.
Team memory for engineers and their AI agents.
LangMem: long-term memory primitives for LangGraph agents that learn and adapt from conversations.
Real-time knowledge graphs that give agents a memory with structure.
30 runnable notebooks teaching everything about agent memory.
LibreAIカタログに掲載されたAIエージェントプロジェクトの中で最もスターの多いもの — honchoが合わなかった場合に検討したい、現実的な候補リストです。
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
カタログの中でhonchoとスター数が近いAIエージェントプロジェクト — 前後の棚もそのまま見て回れます。
A batteries-included browser API so agents can automate the web.
Reef — infrastructure for continually self-improving agents: weight-training and harness-evolving deployments.
Flexible and powerful framework for managing multiple AI agents and handling complex conversations
Put guardrails on LLMs: validate outputs before they reach users.
ZCode — Z.ai's AI coding workspace: desktop app, browser interface and terminal agent in one open-source harness
OpenTelemetry-native observability for LLM apps.