repowiserepowise

MCP server

vectorize-io / hindsight

Hindsight: Agent Memory That Learns

★ 61,199stars on GitHubPythonMIT

This connects Repowise to vectorize-io/hindsight's code. It does not install hindsight.

Repowise maps public repos from their source. Not affiliated with vectorize-io. What is Repowise?

github.com/vectorize-io/hindsightIndexed at 0bd5501

What is hindsight?

Hindsight is an agent memory system for building AI agents that learn over time, not only recall conversation history. Agents store information with retain, search it with recall, and generate disposition-aware responses with reflect. The core is written in Python, with client libraries and tooling in TypeScript, Go and Rust. To install it, run pip install hindsight-api and follow the steps below. Its code health is 7.8 out of 10 (Good), ranked #84 of 4,000+ indexed repos. The last commit to hindsight was on Nov 7, 2026, and 4 people have contributed.

As of commit 0bd5501, indexed Nov 9, 2026. What it is comes from a model reading the README; the rest is measured from the source and git history, not from the maintainer.

For agentshindsight.md

Repo facts: type, license, health, rank, contributors
hindsight facts
TypeMCP server
LicenseMIT
Code health7.8 / 10 (Good)
Health rank#84 of 4,000+ indexed repos
Last commitNov 7, 2026
Contributors4
Bus factor1
Size4,411 files, 654,934 lines

Install hindsight

Commands copied from the README

  1. 1Install the API server from PyPI.
    pip install hindsight-api
  2. 2Set the API key for the LLM provider the server will use.
    export HINDSIGHT_API_LLM_API_KEY=sk-xxx

Releases on GitHub (latest v1.2.3, Nov 19, 2026)

Full install notes on GitHub

Explore hindsight

Ask about this code in your agent

This connects Repowise to hindsight's code. It does not install hindsight. To install hindsight, use the install steps.

Useful if you want to change or fork hindsight: your agent can ask how it is put together and get answers that cite its files.

Ask about hindsight

Answers come from hindsight's README and source, with the files they cite. Two free answers a day without an account.

Featured answer

Hindsight is an agent memory system for building AI agents that learn over time rather than only recalling conversation history. Agents store information with retain, search it with recall, and produce disposition-aware responses with reflect. Requests enter through the API server started by hindsight-api-slim/hindsight_api/main.py, which exposes HTTP routes defined in hindsight-api-slim/hindsight_api/api/http.py. The memory logic lives in hindsight-api-slim/hindsight_api/engine/memory_engine.py, and the hindsight-clients directory provides libraries that let applications use a running server without writing HTTP calls by hand. The hindsight-cli module supplies the hindsight command-line tool, and hindsight-embed runs a local memory server as a background daemon. To start reading, look at main.py and http.py for the API layer, then memory_engine.py for the core logic.

Files per layer in hindsight10 layers.

Basis: Inferred · knowledge graph· as of 0bd5501

Want this for your own codebase? Map your repo free

About this page

Get a page like this for your own codebase

Repowise built this map of vectorize-io/hindsight from its source and git history. It parses every symbol, draws the dependency graph, scores each file's health from complexity, duplication, coverage and churn, finds hotspots and owners in the history, and writes documentation for every module. Your coding agent reads the same index over MCP, so its answers cite real files.

Free for 2 public repos, with a page like this one. No card. See how Repowise works

Prefer local? pip install repowise && repowise init indexes your own checkout, no account needed. Star Repowise on GitHub

vectorize-io/hindsight FAQ

What is Hindsight?

Hindsight is an agent memory system that lets AI agents learn over time rather than only recalling conversation history. It is built to create smarter agents that learn, not just remember.

How does Hindsight work?

Hindsight runs as an API server that agents and applications call to retain, recall and reflect on memories. Client libraries in hindsight-clients connect an application to a running server.

How do I install Hindsight?

Hindsight can be installed with pip install hindsight-api, after which an LLM API key is exported and the hindsight-api command starts the server. Docker and Helm on Kubernetes are also supported. See the install steps

What language is Hindsight written in?

Hindsight is written mainly in Python. Its repository also contains TypeScript, Go, Rust and shell code, including client libraries.

How is the Hindsight code organized?

Hindsight is organized into modules such as hindsight-api-slim for the server, hindsight-clients for client libraries, hindsight-cli for the command-line tool and hindsight-embed for the local daemon. Each module has its own package and manifest.

Who is Hindsight for?

Hindsight is for developers who build AI agents and need a memory layer for them. Operators can run its API server with Docker, Kubernetes or a Python install.

What license does Hindsight use?

Hindsight is released under the MIT license, identified as MIT.

Code health

7.7out of 10Good

Hotspots average 5.1, the place fixes pay off most.

On the leaderboards:#1 of 4 Agent memory layers#19 of 71 Python repos

Key features of hindsight

  • Retain, recall and reflect operations let an agent store information, search memories and generate responses that reflect what it has learned.
  • An LLM Wrapper adds memory to an existing agent by swapping in a wrapped LLM client, so memories are stored and retrieved on every call.
  • The LLM provider is chosen with the HINDSIGHT_API_LLM_PROVIDER setting, covering hosted providers, local models and OpenAI-compatible endpoints.
  • Client libraries are available for Python, Node.js and TypeScript, Go, and a REST API, along with a command-line interface.
  • The API server can run with Docker, on Kubernetes with Helm, or as a plain Python install with pip.

How vectorize-io/hindsight works

Requests enter through the API server, which is started by hindsight-api-slim/hindsight_api/main.py and exposes HTTP routes defined in hindsight-api-slim/hindsight_api/api/http.py. The memory logic sits in hindsight-api-slim/hindsight_api/engine/memory_engine.py, which is one of the most frequently changed files in the code index. The hindsight-clients directory holds the client libraries that let an application store, search and reflect on memories in a running server without writing HTTP calls by hand. The hindsight-cli module provides the hindsight command-line tool, with each subcommand calling the server through one shared API client. The hindsight-embed module runs a local memory server as a background daemon so the hindsight-embed command can work without manual server setup.

The codebase has 4,411 files and 45,337 symbols in 17 modules, led by Python, TypeScript and Go. Code health is 7.7 out of 10, rated good.