vectorize-io/hindsight

What is it?

Key features

• 01Retain, recall and reflect operations for storing and querying agent memories
• 02Recall combines vector, keyword, entity-graph and temporal search with reranking
• 03Memory banks isolate memories per user, agent or project
• 04Built-in MCP endpoint per bank for MCP-compatible clients
• 05Works with 25+ LLM providers, including local models via Ollama and LM Studio
• 06Python, Node.js, Go and CLI clients plus a REST API

Use cases

• Personalizing a chatbot by remembering each user's preferences and history
• Giving coding agents long-term memory of a project's architecture and past sessions
• Adding memory to an existing LLM app with a wrapper around the OpenAI or Anthropic client
• Building autonomous agents that adapt their behavior based on accumulated feedback

Quick start

1. 1

   Start a server with Docker

   docker run -it --pull always --name hindsight --restart unless-stopped -p 8888:8888 -p 9999:9999 -e HINDSIGHT_API_LLM_API_KEY=$OPENAI_API_KEY -v hindsight-data:/home/hindsight/.pg0 ghcr.io/vectorize-io/hindsight:latest
2. 2

   Install the Python client

   pip install hindsight-client -U
3. 3

   Connect the client and retain, recall, or reflect on memories

Things to know

• !Full self-hosting needs an LLM provider API key or a local model setup.
• !Intel Macs need the separate hindsight-all-slim package for embedded use.

Topics

FAQ

What is hindsight?

Agent memory system that stores, recalls and reflects on memories so AI agents learn over time instead of only replaying chat history. Hindsight is an agent memory system aimed at making AI agents that learn, not just remember. Instead of only recalling conversation history, it organizes memories into world facts, the agent's own experiences, consolidated observations, and mental models that are synthesized over time.

Is hindsight free to use? What license does it have?

hindsight is open source under the MIT license. Check the license terms before using it in commercial products.

Is hindsight still maintained?

Actively maintained. The last commit to hindsight was on Oct 20, 2026.

What can I use hindsight for?

Personalizing a chatbot by remembering each user's preferences and history Giving coding agents long-term memory of a project's architecture and past sessions Adding memory to an existing LLM app with a wrapper around the OpenAI or Anthropic client Building autonomous agents that adapt their behavior based on accumulated feedback

How do I install hindsight?

Start a server with Docker: docker run -it --pull always --name hindsight --restart unless-stopped -p 8888:8888 -p 9999:9999 -e HINDSIGHT_API_LLM_API_KEY=$OPENAI_API_KEY -v hindsight-data:/home/hindsight/.pg0 ghcr.io/vectorize-io/hindsight:latest Install the Python client: pip install hindsight-client -U Connect the client and retain, recall, or reflect on memories

Where can I get hindsight?

The source code and installation instructions are on GitHub: https://github.com/vectorize-io/hindsight