marciopuga

marciopuga / cog

Public

Cognitive architecture for Claude Code — persistent memory, self-reflection, and foresight

83
5
100% credibility
Found Mar 26, 2026 at 83 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
AI Summary

Cog provides plain-text conventions and instructions that enable an AI assistant to build, organize, and maintain a persistent multi-tier memory system using simple files.

How It Works

1
🔍 Discover Cog

You hear about Cog, a simple way to give your AI helper a smart, growing memory that remembers your life across chats.

2
📥 Grab the starter files

Download the basic plain-text files and open them in your AI chat tool – it feels like setting up a new notebook.

3
🗣️ Share your world

Chat with your AI about your job, family, hobbies, and goals – it creates cozy spaces to store memories just for you.

4
💬 Chat every day

Talk naturally about events, tasks, people, and ideas – your AI sorts them into hot tips, daily notes, and deep archives automatically.

5
🔍 Spot patterns and insights

Ask your AI to review and reflect – it finds connections, condenses notes, and nudges you with helpful foresight.

6
Keep it fresh
👐
Do it yourself

Occasionally ask for a quick cleanup and refresh to keep everything sharp.

⏰
Set it automatic

Schedule gentle checkups so it stays tidy without you lifting a finger.

🧠 Unlock lasting smarts

Your AI now has a personal knowledge base that grows wiser over time, helping with work, life, and big decisions effortlessly.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 83 to 83 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is cog?

Cog is a plain-text cognitive architecture for Claude Code, turning Anthropic's agentic terminal into a persistent memory system with self-reflection and foresight. You chat with /setup to define personal or work domains—like family health or project tracking—and it scaffolds markdown files that Claude maintains across sessions using Unix tools for search and diffs. No servers or runtime; it's conventions Claude follows, weaving in cognitive ai github patterns like tiered memory (hot priorities, warm domains, cold archives) and tool integrations via MCP for Gmail or calendars.

Why is it gaining traction?

Unlike vector DBs or fragile RAG setups, cog agent github stands out with observable, editable text—grep your insights, git diff evolutions—making memory a debuggable single source of truth. Skills like /reflect condense observations into threads, /foresight nudges strategies, and pipelines (/housekeeping) automate maintenance, hooking devs who want cognitive architectures for language agents without black-box opacity. It's cogito ergo sum for AI: the model reasons over its own git-logged rules.

Who should use this?

Claude Code users tracking work entities (team rosters, action items) or personal habits (health logs, family events) across sessions. AI tinkerers building cognitive architecture github prototypes for agents. Side-project maintainers needing cross-tool context like GitHub PRs or Linear tasks without resetting state.

Verdict

Try it if you're deep in Claude Code—83 stars and 1.0% credibility scream early experiment, but solid docs and zero-code setup make it a low-risk way to bootstrap persistent AI memory. Skip for production; it's a learning tool until pipelines mature.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.