Claude Code Club
Memory & context

10 Repos That Give Claude Code a Second Brain (Memory, Code Graphs, Notes and Plans)

10 minute readUpdated October 2026Explore more

TL;DR

Claude Code forgets everything when a session ends. These 10 free GitHub repos fix that. claude-mem remembers what Claude Code did in past sessions. Understand Anything, CodeGraph and Codebase Memory MCP turn your code into a map Claude Code can search instead of reading file after file. Obsidian Skills lets Claude Code read and write your notes vault. Planning with Files keeps long tasks alive after a crash or /clear. Mem0, Hindsight, OpenViking and Graphiti are bigger memory systems, mostly for people building their own agents and apps. Each section has the GitHub link, live star count, the install steps from the repo's own README, and the catch.

Learn Claude Code. Earn income. Only $9.

👉 https://www.skool.com/claudecodeclub

The 10 repos that give Claude Code a second brain

Star counts below were pulled from GitHub on October 2, 2026, so they will be higher by the time you read this. The list covers four jobs: remember past sessions, map your code, connect your notes and keep long tasks on track. You don't need all 10. The first three items and number 10 plug straight into Claude Code. Mem0, Hindsight, OpenViking and Graphiti need a server or an API key, so start there only if you are building your own agent.

1. claude-mem: Claude Code remembers what it did in past sessions

thedotmack/claude-mem

95,197 stars on October 2, 2026. Apache 2.0 license.

What it does: a persistent memory system built for Claude Code. It captures what Claude does during a session, writes short summaries, and feeds the useful parts back into your next session. You stop re-explaining your project after every restart. It also adds a search tool so Claude can look up past work, like an old bug fix.

How to install it:

bashnpx claude-mem install

Or install it from the plugin marketplace inside Claude Code:

prompt/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

Restart Claude Code. Context from previous sessions will appear in new sessions automatically.

2. Understand Anything: your whole codebase as one clickable map

Egonex-AI/Understand-Anything

85,062 stars on October 2, 2026. MIT license.

What it does: a Claude Code plugin that scans your project with a team of agents and builds a knowledge graph of every file, function, class and dependency. Then it opens an interactive dashboard where you can click any piece and get a plain-English explanation. It also works on docs and note wikis.

How to install it: inside Claude Code, type:

prompt/plugin marketplace add Egonex-AI/Understand-Anything
/plugin install understand-anything

Then analyze your project and open the dashboard:

prompt/understand
/understand-dashboard

Ask questions with `/understand-chat How does the payment flow work?` and re-run `/understand` anytime. It only re-analyzes files that changed.

3. CodeGraph: Claude Code finds the right file instantly instead of searching

colbymchenry/codegraph

72,952 stars on October 2, 2026. MIT license.

What it does: builds a pre-indexed graph of every symbol, call and dependency in your code, and hands it to Claude through an MCP server. Instead of opening files one at a time, Claude asks the graph one question and gets the relevant code back. A file watcher keeps the graph in sync every time you save.

How to install it (Mac or Linux):

bashcurl -fsSL https://raw.githubusercontent.com/colbymchenry/codegraph/main/install.sh | sh

Already have Node? `npm i -g @colbymchenry/codegraph` works too. Then open a new terminal, connect it to your agents, and index each project:

bashcodegraph install
cd your-project
codegraph init

4. Mem0: a memory layer for the agents and apps you build

mem0ai/mem0

66,490 stars on October 2, 2026. Apache 2.0 license.

What it does: a memory layer you add to your own AI apps so they remember users, preferences and past chats. It comes as a Python or JavaScript library, a self-hosted server, or a managed cloud platform.

Easiest way in with Claude Code: install Mem0's skill so Claude knows how to build with it:

bashnpx skills add https://github.com/mem0ai/mem0 --skill mem0

To use the library yourself:

bashpip install mem0ai
# or for JavaScript:
npm install mem0ai

5. Obsidian Skills: Claude Code reads and writes your Obsidian vault

kepano/obsidian-skills

49,091 stars on October 2, 2026. MIT license.

What it does: skills that teach Claude the Obsidian formats, so your notes become a brain Claude can search and update. It covers Obsidian-style Markdown (wikilinks, embeds, callouts, properties), Obsidian Bases, JSON Canvas, the Obsidian CLI, and Defuddle for pulling clean text from web pages.

How to install it: inside Claude Code, type:

prompt/plugin marketplace add kepano/obsidian-skills
/plugin install obsidian@obsidian-skills

Or with the skills installer:

bashnpx skills add https://github.com/kepano/obsidian-skills

Then open Claude Code inside your vault folder and ask it to work with your notes:

promptRead my notes in this vault about my current project, then write a summary note that links back to each one with wikilinks.

6. Codebase Memory MCP: fast answers about huge repos without burning context

DeusData/codebase-memory-mcp

45,693 stars on October 2, 2026. MIT license.

What it does: indexes your codebase into a persistent knowledge graph and gives Claude tools to search it, trace calls, see the architecture and check what a change will affect. The README says it supports 162 languages and ships as a single program with a built-in 3D graph viewer.

How to install it (Mac or Linux):

bashcurl -fsSL https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.sh | bash

Restart Claude Code, then say:

promptIndex this project

To see the graph, run `codebase-memory-mcp --ui=true --port=9749` and open http://localhost:9749 in your browser.

7. Hindsight: agent memory that learns from past runs

vectorize-io/hindsight

44,690 stars on October 2, 2026. MIT license.

What it does: a memory system built to make agents learn over time, not just recall chat history. It sorts memories into facts, experiences and observations, and has three actions: retain, recall and reflect. For coding agents, it builds a memory bank per repo from your git history and past sessions.

How to install it for Claude Code:

bashnpx @vectorize-io/hindsight-coding-agents install claude-code

To run your own Hindsight server (Docker):

bashexport OPENAI_API_KEY=sk-xxx
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

8. OpenViking: one organized place for everything your agent knows

volcengine/OpenViking

39,134 stars on October 2, 2026. AGPL 3.0 license. From Volcengine.

What it does: an open-source context database that stores your agent's knowledge, memories and skills like folders and files. Each folder gets a one-line summary and an overview, so the agent can check what is relevant before reading the full content. Past sessions turn into Markdown memory files you can open and edit.

How to install it: first set up a server. You need uv and Python 3.10 or newer:

bashuv tool install openviking --upgrade && openviking-server init

The setup wizard asks for a model provider. Keep that terminal open, then in a new one install the memory plugin and pick Claude Code:

bashcurl -fsSL https://openviking.ai/install | bash

Restart Claude Code and check it with `/openviking-memory:ov`.

9. Graphiti: knowledge graphs that track how facts change over time

getzep/graphiti

31,387 stars on October 2, 2026. Apache 2.0 license. From Zep.

What it does: a Python framework that builds knowledge graphs from your data in real time. Every fact has a time window, so when something changes, the old fact is marked out of date instead of deleted. You can ask what is true now, or what was true last month.

How to install it: you need Python 3.10+ and a graph database. The README's quickest database option is FalkorDB in Docker:

bashdocker run -p 6379:6379 -p 3000:3000 -it --rm falkordb/falkordb:latest

Then install Graphiti with FalkorDB support:

bashpip install graphiti-core[falkordb]

10. Planning with Files: long jobs survive crashes and pick up where they stopped

OthmanAdi/planning-with-files

27,263 stars on October 2, 2026. MIT license.

What it does: a skill that keeps your plan on disk instead of only in Claude Code's context window. It writes three files in your project: task_plan.md (phases and checkboxes), findings.md (research and decisions) and progress.md (session log). After a /clear, a crash or a new session, Claude reads the plan and resumes at the current phase.

How to install it:

bashnpx skills add OthmanAdi/planning-with-files --skill planning-with-files -g

The README links docs/installation.md for every other install method, including the Claude Code plugin with hooks that re-inject the plan each turn.

Quick start

  1. 1Install claude-mem so Claude remembers your project between sessions.
  2. 2On a big code project, add CodeGraph or Codebase Memory MCP. Both run locally with no API key.
  3. 3If you keep notes in Obsidian, add Obsidian Skills and open Claude Code inside your vault.
  4. 4Before your next long task, install Planning with Files so a crash or /clear never costs you the plan.
Learn Claude Code. Earn income. Only $9.

👉 https://www.skool.com/claudecodeclub

Common questions

  • Are these 10 repos free?

    Yes. All 10 are free and open source. Some need a model or service to run: Mem0, Hindsight and Graphiti default to an OpenAI key (other providers work too), OpenViking needs a model provider for its server, and claude-mem offers an optional hosted observer with a 14-day free trial. CodeGraph and Codebase Memory MCP need no API key.

  • Which one should I install first?

    claude-mem. It is built for Claude Code, installs with one command, and fixes the most common pain: re-explaining your project after every restart.

  • Does adding memory cost me tokens?

    Yes, a little. Any memory that gets loaded into Claude Code's context uses tokens. Tools like claude-mem try to keep that small by loading a short index first and fetching details only when needed. Understand Anything's first scan of a big project is the one that can use a lot.

  • Is my code sent anywhere?

    It depends on the tool. CodeGraph and Codebase Memory MCP say they run 100% locally. claude-mem stores memories locally with an optional cloud sync. Mem0, Hindsight and OpenViking have cloud versions that store data on their servers, plus self-hosted options. Read the README of each tool before you install it.

  • What is the difference between a code graph and a memory tool?

    A code graph (Understand Anything, CodeGraph, Codebase Memory MCP) maps how your code fits together so Claude finds things fast. A memory tool (claude-mem, Mem0, Hindsight) remembers what happened in past conversations. Planning with Files does a third job: it keeps the current task's plan safe on disk.

Keep going

Get more repos like these inside Claude Code Club for $9/mo

Get the other 2 in the memory stack, plus 8,000+ members - $9/mo, cancel anytime.

Join the Club