hindsight
SHA-256Open-source agent memory system with benchmark-leading accuracy; adds learning memory to agents via a simple retain/recall/reflect API, beating typical RAG and knowledge-graph approaches.
Smart Download
Download Download Version
v1.0.0 · 4 MB
Gives AI agents long-term memory that actually learns over time.
Core Features
- Three simple operations: retain, recall, and reflect
- State-of-the-art LongMemEval accuracy, outperforming RAG and knowledge-graph approaches
- Works with 25+ LLM providers, including local models via Ollama for zero API cost
- Flexible deployment: Docker, pip, Helm, embedded no-server mode, or managed cloud
- SDKs for Python, Node.js, and Go, plus a CLI and MCP server
What It Can't Do
- •It is a developer-facing memory backend, not a ready-made chat app—you still need your own agent code. It calls an LLM at runtime, which adds API costs unless you switch to local models via Ollama. Self-hosting relies on PostgreSQL, so you own backups and operations. Some published benchmark scores are self-reported by vendors, so validate on your own workload. It is likely overkill if you only need simple chat history storage.
Use Cases
- Customer support bots that remember user preferences and past issues
- Coding assistants that accumulate project context and personal habits
- Personal assistants that learn schedules, preferences, and relationships over time
- Enterprise knowledge agents that capture team expertise and business context
Hindsight is an open-source agent memory system designed to make AI agents learn over time instead of merely recalling conversation history. It reports state-of-the-art results on the LongMemEval long-term memory benchmark, independently reproduced by researchers at Virginia Tech and The Washington Post. Unlike typical RAG or knowledge-graph setups—and unlike lighter memory tools such as Mem0 or Zep—it exposes a three-operation API (retain, recall, reflect) backed by typed memories, observations, and mental models, giving it benchmark-leading accuracy on long-term memory tasks. Deployment is flexible: a single Docker container, pip install, Helm chart for Kubernetes, an embedded Python mode with no server, or the managed Hindsight Cloud. It supports 25+ LLM providers including OpenAI, Anthropic, Gemini, local models via Ollama, and subscription-based options like Claude Code and GitHub Copilot. Official SDKs cover Python, Node.js/TypeScript, Go, and a CLI, plus an MCP server and an LLM wrapper that adds memory to existing agents in two lines of code. MIT licensed and used in production at Fortune 500 enterprises.
Tags
Getting Started
Download installer
Click the button above to download the installer for your system
Install the software
Double-click the downloaded installer and follow the prompts
Set the OPENAI_API_KEY environment variable to your LLM key
Start the memory server by running the ghcr.io/vectorize-io/hindsight Docker image
Install the Python client with pip install hindsight-client
- Set the OPENAI_API_KEY environment variable to your LLM key
- Start the memory server by running the ghcr.io/vectorize-io/hindsight Docker image
- Install the Python client with pip install hindsight-client
SHA-256 checksum verified
Checksum extracted from GitHub official Release page
SHA256 Checksum
be5833344ac037c7670366936ac8261cbf59f90e5551eee3654ddf4b73dec7e8This checksum is extracted from the GitHub Release page. Verify file integrity after download.
All SHA-256 checksums on this platform are extracted from the project's official GitHub Release page, without any modification. You can independently verify them on the GitHub Releases page.
Open Source Transparency
View GitHub SourceUninstall Info
For Docker: run docker stop hindsight && docker rm hindsight, then docker volume rm hindsight-data to wipe all stored memories. For pip installs: pip uninstall hindsight-api hindsight-client. For Helm: helm uninstall hindsight and clean up the associated PVCs. The managed cloud version requires no uninstall—just delete your data from the dashboard.
No Extra Dependencies
Ready to use after download. No additional runtime required.
Having issues? Check the FAQ below
5 FAQs
Similar Projects
mempalace
Open-source local-first AI memory system with verbatim storage and semantic search — zero API calls, no keys, top benchmark scores.
ollama
Ollama lets you download, run, and manage large language models locally. One command, multiple platforms, endless possibilities.
llama.cpp
High-performance LLM inference engine in C/C++ with minimal dependencies, supporting quantized models (1.5–8 bit) and diverse hardware (Apple Silicon, CUDA, Vulkan, etc.).